Coworker — agente

Tienda de herramientas

Las categorías Núcleo vienen listas para usar; el resto es opt-in. Elige un workspace registrado y guarda para persistir en la base de datos local.

grid_view 26 categorías build 157 herramientas toggle_on 15 activas (vista actual)

Las etiquetas «Configurado en Coworker» solo indican claves presentes en Variables de entorno y secretos (BD local), no variables del proceso del servidor ni archivos .env.

Filtros

Categoría

Vista compacta

Agente (persistencia en BD)

Gestionar agentes

Preferencias de herramientas

Mostrar snapshot tools

Variables globales (LLM y runtime)

En el repositorio, consulta agent/README.md para la tabla completa de variables.

Proveedor LLM

AGENT_LLM_PROVIDER Config

Default: openai

openai | anthropic | ollama | lmstudio | airllm.
AGENT_LLM_MAX_TOOLS Config

Default: 128

Máx. herramientas por petición (API OpenAI-compatible). LM Studio limita a 128; 0 = sin recorte.
OPENAI_API_KEY Secreto

Clave API OpenAI (o compatible).
OPENAI_BASE_URL · OPENAI_MODEL Config

Default: https://api.openai.com/v1 y gpt-4o-mini

Base URL y modelo si usas OpenAI o un proxy compatible.
ANTHROPIC_API_KEY Secreto

Clave API de Anthropic.
ANTHROPIC_MODEL Config

Modelo Claude (default en README si no lo defines).
OLLAMA_BASE_URL · OLLAMA_MODEL Config

Default: http://127.0.0.1:11434 y llama3.2

Instancia local Ollama.
AIRLLM_MODEL Config

ID Hugging Face o ruta local del modelo. Ver AIRLLM_COMPRESSION, AIRLLM_MAX_* en README.
AIRLLM_HF_TOKEN Secreto

Cuándo: Si el modelo es gated en Hugging Face

Token de lectura HF.

Registry del coworker (CLI / web)

AGENT_REGISTRY_INCLUDE_PLAN Feature flag

Default: 1

Registrar plan agent.
AGENT_REGISTRY_INCLUDE_ASK_USER Feature flag

Default: 1

Registrar pregunta al usuario.
AGENT_REGISTRY_INCLUDE_DELEGATE Feature flag

Default: 1

Registrar delegate_subagent y aliases si hay proveedor LLM.

GUI web y sesión

AGENT_WEB_SECRET Secreto

Default: change-me… (solo desarrollo)

Firma de cookies de sesión; obligatorio cambiar en producción.
AGENT_WEB_WORKSPACE Config

Ruta por defecto del workspace en formularios.

Incluidas (núcleo)

Listas para el día a día del agente

Core Núcleo

Edición del workspace

Lectura, listado, parches y escritura de archivos bajo la raíz del proyecto.

Herramientas activas 6/6

tune Configurar variables en Ajustes

Entorno / secretos

CONFIRM_WRITES Config Falta en BD

Default: 0 (desactivado)

Si es 1/true, write_file/apply_patch/generate_csv generan cambios pendientes; generate_pptx falla (binario).

apply_patch

Apply a unified diff patch to a file inside the workspace.

Esquema

{
  "properties": {
    "max_patch_bytes": {
      "description": "Max allowed bytes for patch",
      "type": "integer"
    },
    "patch": {
      "description": "Unified diff string",
      "type": "string"
    },
    "path": {
      "description": "Relative path of target file (optional if patch includes +++ b/path)",
      "type": "string"
    }
  },
  "required": [
    "patch"
  ],
  "type": "object"
}

generate_csv

Write a CSV file under the workspace (relative path). rows: array of arrays (row values) or array of objects (dict keys become columns). Optional headers for array-of-arrays mode, or column order for object rows. Set excel_utf8 true for UTF-8 BOM so Excel opens encoding correctly.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "delimiter": {
      "description": "Single character, default comma",
      "type": "string"
    },
    "excel_utf8": {
      "description": "If true, prefix UTF-8 BOM for Excel",
      "type": "boolean"
    },
    "headers": {
      "description": "Column names: prepended as first row for array rows, or column order for object rows",
      "items": {
        "type": "string"
      },
      "type": "array"
    },
    "max_bytes": {
      "description": "Max UTF-8 size (default 200000)",
      "type": "integer"
    },
    "path": {
      "description": "Relative path under workspace (e.g. out/data.csv)",
      "type": "string"
    },
    "rows": {
      "description": "Array of row arrays, or array of objects with string keys",
      "items": {
        "anyOf": [
          {
            "items": {
              "anyOf": [
                {
                  "type": "string"
                },
                {
                  "type": "number"
                },
                {
                  "type": "boolean"
                },
                {
                  "type": "null"
                }
              ]
            },
            "type": "array"
          },
          {
            "additionalProperties": {
              "anyOf": [
                {
                  "type": "string"
                },
                {
                  "type": "number"
                },
                {
                  "type": "boolean"
                },
                {
                  "type": "null"
                }
              ]
            },
            "type": "object"
          }
        ]
      },
      "type": "array"
    }
  },
  "required": [
    "path",
    "rows"
  ],
  "type": "object"
}

generate_pptx

Create a PowerPoint (.pptx) under the workspace. Optional presentation_title adds a title slide. slides: array of {title, bullets? (string array), body? (single paragraph)}. Not available when CONFIRM_WRITES is enabled unless approval bypass is active (binary output).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "max_bytes": {
      "description": "Max file size (default 2000000)",
      "type": "integer"
    },
    "path": {
      "description": "Relative path; must end with .pptx",
      "type": "string"
    },
    "presentation_title": {
      "description": "Optional main title slide text",
      "type": "string"
    },
    "slides": {
      "description": "Content slides with title and bullets and/or body",
      "items": {
        "properties": {
          "body": {
            "type": "string"
          },
          "bullets": {
            "items": {
              "type": "string"
            },
            "type": "array"
          },
          "title": {
            "type": "string"
          }
        },
        "type": "object"
      },
      "type": "array"
    }
  },
  "required": [
    "path"
  ],
  "type": "object"
}

list_files

List files under a directory (relative to workspace). Useful to discover what to read/edit.

Esquema

{
  "properties": {
    "max_results": {
      "description": "Max number of file paths to return",
      "type": "integer"
    },
    "path": {
      "description": "Relative directory; empty = workspace root",
      "type": "string"
    },
    "pattern": {
      "description": "Simple glob over filename, e.g. *.py",
      "type": "string"
    }
  },
  "required": [],
  "type": "object"
}

read_file

Read a UTF-8 text file inside the workspace (relative path).

Esquema

{
  "properties": {
    "max_bytes": {
      "description": "Max bytes to read (default 200000)",
      "type": "integer"
    },
    "path": {
      "description": "Path relative to workspace root",
      "type": "string"
    }
  },
  "required": [
    "path"
  ],
  "type": "object"
}

write_file

Create or overwrite a text file inside the workspace (relative path).

Esquema

{
  "properties": {
    "content": {
      "description": "File content as UTF-8 text",
      "type": "string"
    },
    "max_bytes": {
      "description": "Max allowed bytes for content",
      "type": "integer"
    },
    "path": {
      "description": "Relative path under workspace root",
      "type": "string"
    }
  },
  "required": [
    "path",
    "content"
  ],
  "type": "object"
}

Discovery Núcleo

Búsqueda

Grep en el repo y descubrimiento de herramientas disponibles.

Herramientas activas 4/4

grep

Search files with a Python regex under a path (default: whole workspace).

Esquema

{
  "properties": {
    "max_results": {
      "type": "integer"
    },
    "path": {
      "description": "Relative directory or file; empty = workspace root",
      "type": "string"
    },
    "pattern": {
      "type": "string"
    }
  },
  "required": [
    "pattern"
  ],
  "type": "object"
}

parse_test_failures

Parse test runner stdout/stderr into structured failures (pytest, jest, unittest). Pass raw text from run_terminal or pasted logs; helps when UI truncates output. format: auto (detect), pytest, jest, or unittest.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "format": {
      "description": "auto detects runner",
      "enum": [
        "auto",
        "pytest",
        "jest",
        "unittest"
      ],
      "type": "string"
    },
    "output": {
      "description": "Raw test output",
      "type": "string"
    }
  },
  "required": [
    "output"
  ],
  "type": "object"
}

tool_search

Search tools by substring in name/description (current registry).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "limit": {
      "type": "integer"
    },
    "q": {
      "type": "string"
    },
    "query": {
      "type": "string"
    }
  },
  "required": [],
  "type": "object"
}

ToolSearchTool

Search tools by substring in name/description (current registry).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "limit": {
      "type": "integer"
    },
    "q": {
      "type": "string"
    },
    "query": {
      "type": "string"
    }
  },
  "required": [],
  "type": "object"
}

Verification Núcleo

Terminal

Ejecución de comandos con allowlist (tests, git vía argv, etc.).

Herramientas activas 2/2

tune Configurar variables en Ajustes

Entorno / secretos

AGENT_EXEC_ALLOWLIST Config Falta en BD

Default: git,python,py,pytest,pip,dir,echo,type,npm,npx

Lista de ejecutables permitidos para run_terminal (separados por coma).
AGENT_EXEC_ALLOWLIST_EXTRA Config Falta en BD

Allowlist adicional (p. ej. ampliada desde la GUI en sesión).
AGENT_TERMINAL_CONFIRM_FULL Config Falta en BD

Si está activo, confirma por argv completo además del stem del ejecutable.
AGENT_DOCKER_TERMINAL · AGENT_DOCKER_IMAGE Config Falta en BD

Cuándo: Solo si ejecutas la terminal dentro de Docker

Con AGENT_DOCKER_TERMINAL=1, run_terminal usa docker run; define AGENT_DOCKER_IMAGE y opcionalmente AGENT_DOCKER_RUN_ARGS, AGENT_DOCKER_BIN.

run_terminal

Run a command with argv list (no shell). Uses allowlist for argv[0].

Esquema

{
  "properties": {
    "argv": {
      "items": {
        "type": "string"
      },
      "type": "array"
    },
    "cwd": {
      "description": "Relative cwd under workspace",
      "type": "string"
    },
    "timeout_sec": {
      "type": "number"
    }
  },
  "required": [
    "argv"
  ],
  "type": "object"
}

verify_workspace

Run predefined verify profiles under the workspace (pytest, pytest_quiet, ruff_check, ruff_format_check). Same allowlist/Docker policy as run_terminal. Structured JSON (schema_version). For ruff profiles, allow the ruff stem via env AGENT_EXEC_ALLOWLIST_EXTRA, .coworker-allowlist, or GUI session allowlist. Env: VERIFY_MAX_CAPTURE_BYTES, VERIFY_PYTEST_INNER_ARGS.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "cwd": {
      "description": "Optional relative cwd under workspace",
      "type": "string"
    },
    "profile": {
      "description": "pytest, pytest_quiet, ruff_check, or ruff_format_check",
      "enum": [
        "pytest",
        "pytest_quiet",
        "ruff_check",
        "ruff_format_check"
      ],
      "type": "string"
    },
    "targets": {
      "description": "Relative paths; empty uses tests/ if present else .",
      "items": {
        "type": "string"
      },
      "type": "array"
    },
    "timeout_sec": {
      "description": "Optional timeout (default 300 tests / 120 lint)",
      "type": "number"
    }
  },
  "required": [
    "profile"
  ],
  "type": "object"
}

Tasks Núcleo

TODO en agente

Sincronización del tablero TODO con el modelo.

Herramientas activas 2/2

todo_write

Update the session todo list (merge or replace). Use to track multi-step work. Same semantics as Cursor TodoWriteTool.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "merge": {
      "description": "If true, merge with existing todos; if false, replace the list.",
      "type": "boolean"
    },
    "todos": {
      "items": {
        "additionalProperties": true,
        "properties": {
          "content": {
            "type": "string"
          },
          "id": {
            "type": "string"
          },
          "status": {
            "description": "e.g. pending, in_progress, completed, cancelled",
            "type": "string"
          }
        },
        "required": [
          "content"
        ],
        "type": "object"
      },
      "type": "array"
    }
  },
  "required": [
    "todos"
  ],
  "type": "object"
}

TodoWriteTool

Update the session todo list (merge or replace). Use to track multi-step work. Same semantics as Cursor TodoWriteTool.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "merge": {
      "description": "If true, merge with existing todos; if false, replace the list.",
      "type": "boolean"
    },
    "todos": {
      "items": {
        "additionalProperties": true,
        "properties": {
          "content": {
            "type": "string"
          },
          "id": {
            "type": "string"
          },
          "status": {
            "description": "e.g. pending, in_progress, completed, cancelled",
            "type": "string"
          }
        },
        "required": [
          "content"
        ],
        "type": "object"
      },
      "type": "array"
    }
  },
  "required": [
    "todos"
  ],
  "type": "object"
}

Human Núcleo

Preguntas al usuario

Pausa el flujo para que el humano responda en la UI.

Herramientas activas 1/1

tune Configurar variables en Ajustes

Entorno / secretos

AGENT_REGISTRY_INCLUDE_ASK_USER Flag Falta en BD

Default: 1

Incluir snap_ask_user_question en el registry.

snap_ask_user_question

Emit a structured request to ask the human user a question (to be handled by the frontend/orchestrator).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "allow_multiple": {
      "description": "Si true, permite selecci\u00f3n m\u00faltiple sobre options.",
      "type": "boolean"
    },
    "context": {
      "description": "Contexto adicional sobre por qu\u00e9 se hace la pregunta.",
      "type": "string"
    },
    "options": {
      "description": "Opciones de respuesta sugeridas (opcional).",
      "items": {
        "type": "string"
      },
      "type": "array"
    },
    "question": {
      "description": "Pregunta que se quiere hacer al usuario humano.",
      "type": "string"
    }
  },
  "required": [
    "question"
  ],
  "type": "object"
}

Más herramientas

Extensiones e integraciones opcionales

VCS Extensión

Git

Estado, diff, log, ramas, merge y operaciones de repositorio.

Herramientas activas 0/16

tune Configurar variables en Ajustes

Entorno / secretos

ENABLE_GIT_TOOLS Flag Falta en BD

Default: 1 (activo)

Pon 0/false para no registrar las herramientas git_*.

Jupyter Extensión

Notebooks

Edición de celdas en notebooks (.ipynb).

Herramientas activas 0/2

notebook_edit

Edit or create a Jupyter notebook under the workspace. Use cell_idx (0-based), new_source, is_new_cell, optional cell_type (code|markdown).

Esquema

{
  "additionalProperties": true,
  "properties": {
    "cell_idx": {
      "description": "Cell index (0-based)",
      "type": "integer"
    },
    "cell_type": {
      "enum": [
        "code",
        "markdown"
      ],
      "type": "string"
    },
    "is_new_cell": {
      "description": "Insert new cell at cell_idx",
      "type": "boolean"
    },
    "new_source": {
      "description": "New cell source",
      "type": "string"
    },
    "path": {
      "description": "Relative path to .ipynb",
      "type": "string"
    }
  },
  "required": [
    "path",
    "new_source"
  ],
  "type": "object"
}

NotebookEditTool

Edit or create a Jupyter notebook under the workspace. Use cell_idx (0-based), new_source, is_new_cell, optional cell_type (code|markdown).

Esquema

{
  "additionalProperties": true,
  "properties": {
    "cell_idx": {
      "description": "Cell index (0-based)",
      "type": "integer"
    },
    "cell_type": {
      "enum": [
        "code",
        "markdown"
      ],
      "type": "string"
    },
    "is_new_cell": {
      "description": "Insert new cell at cell_idx",
      "type": "boolean"
    },
    "new_source": {
      "description": "New cell source",
      "type": "string"
    },
    "path": {
      "description": "Relative path to .ipynb",
      "type": "string"
    }
  },
  "required": [
    "path",
    "new_source"
  ],
  "type": "object"
}

Async Extensión

Tareas en background

Herramientas tipo Task* (crear, listar, salida, parar…).

Herramientas activas 0/9

tune Configurar variables en Ajustes

Entorno / secretos

Las tareas en background (Task*, autostart) necesitan un proveedor LLM disponible en el registry al construirlo (mismo requisito que el chat). Revisa el bloque «Variables globales» arriba.

snap_cron_create

Register a logical cron entry for this agent run (5-field cron schedule + command description). Does not install system crontab; use for planning and tracking recurring work.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "action": {
      "description": "Alias for command if command is omitted.",
      "type": "string"
    },
    "command": {
      "description": "What to run or do when the schedule fires (natural language or shell-ish command text).",
      "type": "string"
    },
    "enabled": {
      "description": "Whether the entry is active (default true).",
      "type": "boolean"
    },
    "metadata": {
      "additionalProperties": true,
      "description": "Optional extra metadata.",
      "type": "object"
    },
    "name": {
      "description": "Human-readable name for the schedule.",
      "type": "string"
    },
    "schedule": {
      "description": "Cron expression: minute hour day-of-month month day-of-week (5 fields, space-separated).",
      "type": "string"
    },
    "timezone": {
      "description": "Optional IANA or label timezone hint.",
      "type": "string"
    }
  },
  "required": [
    "name",
    "schedule"
  ],
  "type": "object"
}

snap_cron_delete

Remove a cron entry from this run by cron_id (from snap_cron_list).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "cron_id": {
      "description": "Identifier returned when the cron was created.",
      "type": "string"
    },
    "id": {
      "description": "Alias for cron_id.",
      "type": "string"
    }
  },
  "type": "object"
}

snap_cron_list

List cron entries registered for this agent run (see snap_cron_create).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "limit": {
      "description": "Max entries to return (default 50).",
      "type": "integer"
    }
  },
  "type": "object"
}

snap_task_create

Create a tracked task. Set autostart=true to run it in the background (single slot per run); requires LLM. Uses the same sub-agent engine as delegate_subagent.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "autostart": {
      "description": "If true, start execution immediately in a background thread (one at a time per run).",
      "type": "boolean"
    },
    "description": {
      "description": "Optional task description.",
      "type": "string"
    },
    "metadata": {
      "additionalProperties": true,
      "description": "Optional arbitrary metadata.",
      "type": "object"
    },
    "title": {
      "description": "Short title for the task.",
      "type": "string"
    }
  },
  "required": [
    "title"
  ],
  "type": "object"
}

snap_task_get

Fetch one task by id (status, metadata, summary fields).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "task_id": {
      "description": "Identifier of the task to fetch.",
      "type": "string"
    }
  },
  "required": [
    "task_id"
  ],
  "type": "object"
}

snap_task_list

List tasks for this session run.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "cursor": {
      "description": "Reserved for pagination (unused).",
      "type": "string"
    },
    "limit": {
      "description": "Optional max number of tasks.",
      "type": "integer"
    },
    "status": {
      "description": "Optional status filter.",
      "type": "string"
    }
  },
  "type": "object"
}

snap_task_output

Read incremental streamed output from a task. since_seq is the number of log entries already consumed (index).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "since_seq": {
      "description": "Index offset into the output log (entries already read).",
      "type": "integer"
    },
    "task_id": {
      "description": "Identifier of the task.",
      "type": "string"
    }
  },
  "required": [
    "task_id"
  ],
  "type": "object"
}

snap_task_stop

Request cancellation of a running task (best-effort via cancel_check).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "reason": {
      "description": "Optional stop reason.",
      "type": "string"
    },
    "task_id": {
      "description": "Identifier of the task to stop.",
      "type": "string"
    }
  },
  "required": [
    "task_id"
  ],
  "type": "object"
}

snap_task_update

Update task fields; set start=true to begin execution if not running.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "description": {
      "description": "Optional updated description.",
      "type": "string"
    },
    "metadata": {
      "additionalProperties": true,
      "description": "Optional metadata patch.",
      "type": "object"
    },
    "start": {
      "description": "If true, start the task in background (requires LLM; respects single slot).",
      "type": "boolean"
    },
    "status": {
      "description": "Optional updated status.",
      "type": "string"
    },
    "task_id": {
      "description": "Identifier of the task to update.",
      "type": "string"
    },
    "title": {
      "description": "Optional updated title.",
      "type": "string"
    }
  },
  "required": [
    "task_id"
  ],
  "type": "object"
}

Session Extensión

Plan y worktrees

Modo plan, worktrees de git y override de workspace por conversación.

Herramientas activas 0/4

snap_plan_enter

Set preferred chat mode to plan for this conversation (next messages use read-only tools in UI).

Esquema

{
  "additionalProperties": true,
  "properties": {
    "branch": {
      "type": "string"
    },
    "note": {
      "type": "string"
    },
    "path": {
      "type": "string"
    },
    "worktree_path": {
      "type": "string"
    }
  },
  "type": "object"
}

snap_plan_exit

Clear preferred plan mode for this conversation.

Esquema

{
  "additionalProperties": true,
  "properties": {
    "branch": {
      "type": "string"
    },
    "note": {
      "type": "string"
    },
    "path": {
      "type": "string"
    },
    "worktree_path": {
      "type": "string"
    }
  },
  "type": "object"
}

snap_worktree_enter

git worktree add under workspace; set effective workspace for this conversation to the new tree.

Esquema

{
  "additionalProperties": true,
  "properties": {
    "branch": {
      "type": "string"
    },
    "note": {
      "type": "string"
    },
    "path": {
      "type": "string"
    },
    "worktree_path": {
      "type": "string"
    }
  },
  "type": "object"
}

snap_worktree_exit

Clear workspace override; optionally git worktree remove if path given.

Esquema

{
  "additionalProperties": true,
  "properties": {
    "branch": {
      "type": "string"
    },
    "note": {
      "type": "string"
    },
    "path": {
      "type": "string"
    },
    "worktree_path": {
      "type": "string"
    }
  },
  "type": "object"
}

Analysis Extensión

LSP e IDE

Language Server, esquemas y contexto de símbolos (solo lectura).

Herramientas activas 0/6

tune Configurar variables en Ajustes

Entorno / secretos

AGENT_LSP_COMMAND Config Falta en BD

Comando del language server (tokens separados por espacio).
AGENT_LSP_COMMAND_PYTHON Config Falta en BD

Comando LSP solo para language_id=python (si no, se intenta pylsp).

LSPTool

LSP via stdio (AGENT_LSP_COMMAND or pylsp for python). Actions: document_symbol, hover (needs line/character).

Esquema

{
  "additionalProperties": true,
  "properties": {
    "action": {
      "description": "document_symbol | hover",
      "enum": [
        "document_symbol",
        "document_symbols",
        "symbols",
        "hover"
      ],
      "type": "string"
    },
    "character": {
      "type": "integer"
    },
    "language_id": {
      "description": "e.g. python, typescript",
      "type": "string"
    },
    "line": {
      "type": "integer"
    },
    "path": {
      "description": "Relative file path under workspace",
      "type": "string"
    }
  },
  "required": [
    "path"
  ],
  "type": "object"
}

snap_lsp_symbol_context

Symbol + hover + nearby lines via LSP (AGENT_LSP_COMMAND / pylsp).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "character": {
      "type": "integer"
    },
    "language_id": {
      "type": "string"
    },
    "line": {
      "type": "integer"
    },
    "name": {
      "type": "string"
    },
    "uri": {
      "type": "string"
    }
  },
  "required": [
    "uri",
    "line",
    "character"
  ],
  "type": "object"
}

snap_lsp_tool

LSP via stdio (AGENT_LSP_COMMAND or pylsp for python). Actions: document_symbol, hover (needs line/character).

Esquema

{
  "additionalProperties": true,
  "properties": {
    "action": {
      "description": "document_symbol | hover",
      "enum": [
        "document_symbol",
        "document_symbols",
        "symbols",
        "hover"
      ],
      "type": "string"
    },
    "character": {
      "type": "integer"
    },
    "language_id": {
      "description": "e.g. python, typescript",
      "type": "string"
    },
    "line": {
      "type": "integer"
    },
    "path": {
      "description": "Relative file path under workspace",
      "type": "string"
    }
  },
  "required": [
    "path"
  ],
  "type": "object"
}

snapshot_classifyForCollapse

MCP collapse helper (stub).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "content": {
      "type": "string"
    },
    "metadata": {
      "type": "object"
    },
    "tool_name": {
      "type": "string"
    }
  },
  "required": [
    "tool_name",
    "content"
  ],
  "type": "object"
}

snapshot_formatters

LSP formatting mirror (not supported server-side in this port).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "language_id": {
      "type": "string"
    },
    "options": {
      "type": "object"
    },
    "range": {
      "type": "object"
    },
    "uri": {
      "type": "string"
    }
  },
  "required": [
    "uri",
    "language_id"
  ],
  "type": "object"
}

snapshot_schemas

Schema mirror (not supported).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "operation": {
      "type": "string"
    },
    "payload": {
      "type": "object"
    }
  },
  "required": [
    "operation",
    "payload"
  ],
  "type": "object"
}

Network Extensión

Web

Búsqueda, fetch HTTP y TTS opcional (snap_tts_synthesize con ENABLE_TTS_TOOLS).

Herramientas activas 0/2

tune Configurar variables en Ajustes

Entorno / secretos

snap_web_search usa DuckDuckGo (JSON y, si hace falta, resultados HTML con snippet); snap_web_fetch obtiene HTML sin ejecutar JavaScript. Precios en sitios de billetes suelen cargarse en el navegador: para cotización fiable hace falta MCP con navegador o API del operador. snap_tts_synthesize requiere ENABLE_TTS_TOOLS=1; OpenAI usa OPENAI_API_KEY; Edge requiere el grupo Poetry «tts» (edge-tts).

ENABLE_TTS_TOOLS Flag Falta en BD

Default: vacío (desactivado)

1/true para registrar snap_tts_synthesize (texto → audio bajo el workspace).
OPENAI_API_KEY · OPENAI_BASE_URL Secreto Falta en BD

Cuándo: Proveedor OpenAI TTS

Clave y base URL (p. ej. https://api.openai.com/v1) para provider=openai.
AGENT_TTS_PROVIDER Config Falta en BD

openai | edge si no pasas provider en la tool; por defecto openai si hay OPENAI_API_KEY, si no edge.
AGENT_TTS_OPENAI_MODEL Config Falta en BD

Default: tts-1

Modelo OpenAI para audio/speech (p. ej. tts-1, tts-1-hd).
Config

Cuándo: Proveedor Edge TTS

Instala dependencias con poetry install --with tts (paquete edge-tts) para provider=edge sin API key.

snap_web_fetch

Fetch raw text content from an HTTP(S) URL with a size limit.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "max_bytes": {
      "description": "Maximum number of bytes to read from the response body (default 100000).",
      "type": "integer"
    },
    "url": {
      "description": "HTTP(S) URL to fetch.",
      "type": "string"
    }
  },
  "required": [
    "url"
  ],
  "type": "object"
}

snap_web_search

Web search via DuckDuckGo: instant-answer JSON first, then organic HTML results (title, url, snippet) if the JSON response is sparse. For travel/prices, use snap_web_fetch on selected URLs; many booking sites load fares in the browser (JS), so HTML may lack live prices.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "max_results": {
      "description": "Maximum number of results to return (default 5, max 20).",
      "type": "integer"
    },
    "query": {
      "description": "Search query (short natural language string).",
      "type": "string"
    }
  },
  "required": [
    "query"
  ],
  "type": "object"
}

Integration Extensión

MCP

Model Context Protocol: ejecutar, listar y leer recursos.

Herramientas activas 0/3

tune Configurar variables en Ajustes

Entorno / secretos

AGENT_MCP_SERVERS Config Falta en BD

JSON con servidores MCP stdio; registra tools mcp_<id>_*.
AGENT_MCP_CALL_TIMEOUT_SEC Config Falta en BD

Default: 120

Timeout por llamada a tools MCP embebidas y fallback stdio de snap_mcp_execute.
SNAP_MCP_BASE_URL Config Falta en BD

Cuándo: Opcional

Si está definida, las snap_mcp_* usan backend HTTP Snap en lugar de solo stdio.
SNAP_MCP_API_KEY Secreto Falta en BD

Cuándo: Si usas SNAP_MCP_BASE_URL

Bearer opcional para el backend Snap MCP HTTP.
SNAP_MCP_TIMEOUT_SEC Config Falta en BD

Default: 30

Timeout HTTP hacia el backend Snap MCP.

snap_list_mcp_resources

List MCP resources known by Snap backend.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "kind": {
      "description": "Optional resource kind filter.",
      "type": "string"
    },
    "server_id": {
      "description": "Optional MCP server identifier to filter by.",
      "type": "string"
    }
  },
  "required": [],
  "type": "object"
}

snap_mcp_execute

Execute an MCP tool on a given server. If SNAP_MCP_BASE_URL is set, uses the Snap HTTP backend; otherwise runs the configured stdio MCP server in-process (same sources as AGENT_MCP_SERVERS).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "arguments": {
      "additionalProperties": true,
      "description": "Arbitrary JSON arguments to pass to the MCP tool.",
      "type": "object"
    },
    "server_id": {
      "description": "Identifier of the MCP server to target.",
      "type": "string"
    },
    "timeout_sec": {
      "description": "Optional timeout in seconds for the backend call.",
      "type": "number"
    },
    "tool_name": {
      "description": "Name of the MCP tool to invoke.",
      "type": "string"
    }
  },
  "required": [
    "server_id",
    "tool_name"
  ],
  "type": "object"
}

snap_read_mcp_resource

Read a specific MCP resource via Snap backend.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "server_id": {
      "description": "Optional MCP server identifier to disambiguate.",
      "type": "string"
    },
    "uri": {
      "description": "Resource URI to read (MCP resource identifier).",
      "type": "string"
    }
  },
  "required": [
    "uri"
  ],
  "type": "object"
}

Remote Extensión

SSH y red

SSH vía Paramiko, Wake-on-LAN, escaneo TCP de puertos (opt-in) u operaciones remotas similares.

Herramientas activas 0/2

tune Configurar variables en Ajustes

Entorno / secretos

ENABLE_NETWORK_SCAN_TOOLS Flag Falta en BD

Default: vacío (desactivado)

1/true para registrar snap_tcp_port_scan (escaneo TCP / estilo nmap -sT).
NETWORK_SCAN_ALLOW_NMAP Flag Falta en BD

Default: vacío (desactivado)

1/true para permitir backend=nmap con argv fijo (requiere nmap en PATH).
NETWORK_SCAN_HOST_ALLOWLIST Config Falta en BD

Opcional: hosts o CIDRs separados por comas; si está definido, solo esos destinos.

snap_paramiko_ssh

Open an SSH connection with Paramiko and optionally run one remote command. Authenticate with password and/or private key file. You may set username/password directly or use username_binding / password_binding to read values from Settings → Env vars y secrets (browser session). If trust_host_key=true, known_hosts is not loaded first (any host key accepted); the new key is still appended to ~/.ssh/known_hosts after connect.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "command": {
      "description": "Remote shell command to run. If empty, only tests the connection (runs a no-op shell check).",
      "type": "string"
    },
    "host": {
      "description": "SSH server hostname or IP.",
      "type": "string"
    },
    "key_passphrase": {
      "description": "Passphrase for the private key, if encrypted.",
      "type": "string"
    },
    "key_path": {
      "description": "Path to private key file (optional if password is set).",
      "type": "string"
    },
    "max_output_bytes": {
      "description": "Max combined stdout+stderr bytes to return (default 256000).",
      "type": "integer"
    },
    "password": {
      "description": "Password (optional if key_path is set).",
      "type": "string"
    },
    "password_binding": {
      "description": "Session binding key for password. Used only when password is empty (optional if key_path is set).",
      "type": "string"
    },
    "port": {
      "description": "SSH port (default 22).",
      "type": "integer"
    },
    "timeout_sec": {
      "description": "Connection and channel timeout in seconds (default 30).",
      "type": "number"
    },
    "trust_host_key": {
      "description": "If false (default): load ~/.ssh/known_hosts, verify known hosts, and automatically add new hosts to that file. If true: do not load known_hosts first (accept any host key), then append this server\u0027s key to ~/.ssh/known_hosts after connect.",
      "type": "boolean"
    },
    "username": {
      "description": "SSH username (optional if username_binding is set).",
      "type": "string"
    },
    "username_binding": {
      "description": "Session binding key for username (secrets override env). Used only when username is empty.",
      "type": "string"
    }
  },
  "required": [
    "host"
  ],
  "type": "object"
}

snap_wake_on_lan

Send a Wake-on-LAN magic packet over UDP. Target must have WoL enabled. Use the subnet broadcast address (e.g. 192.168.1.255) if 255.255.255.255 does not reach the host. Port 9 (discard) is typical; port 7 (echo) is also used.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "broadcast_address": {
      "description": "UDP destination IP (default 255.255.255.255).",
      "type": "string"
    },
    "count": {
      "description": "How many times to send the same packet (default 3, max 5).",
      "type": "integer"
    },
    "mac": {
      "description": "MAC address, e.g. aa:bb:cc:dd:ee:ff or aa-bb-cc-dd-ee-ff.",
      "type": "string"
    },
    "port": {
      "description": "UDP port (default 9).",
      "type": "integer"
    }
  },
  "required": [
    "mac"
  ],
  "type": "object"
}

Protocols Extensión

Protocolos de red

FTP/FTPS, SFTP, SMB/CIFS y LDAP/LDAPS. Transferencia de archivos e integración con directorios empresariales.

Herramientas activas 0/15

tune Configurar variables en Ajustes

Entorno / secretos

ENABLE_SMB_TOOLS Flag Falta en BD

Default: vacío (desactivado)

Debe ser 1/true para registrar smb_* (requiere impacket instalado).

Dorking Extensión

Google Hacking

Dorks personalizados, enumeración de índice (site:), archivos expuestos (filetype:) y presets. Activado por defecto; desactivar con ENABLE_GHACK_TOOLS=0.

Herramientas activas 0/4

tune Configurar variables en Ajustes

Entorno / secretos

ENABLE_GHACK_TOOLS Flag Falta en BD

Default: 1 (activo)

0/false desactiva ghack_*; por defecto registradas (dorks vía DuckDuckGo HTML o SerpAPI).
SERPAPI_KEY Secreto Falta en BD

Cuándo: engine=serpapi

Clave API de SerpAPI para usar el motor Google (mayor cobertura que DuckDuckGo).

ghack_dork

Ejecuta una query de Google dork / búsqueda avanzada. Motor 'duckduckgo' (sin clave) o 'serpapi' (requiere SERPAPI_KEY). Soporta operadores: site:, filetype:, inurl:, intitle:, intext:, etc.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "engine": {
      "description": "Motor de b\u00fasqueda: \u0027duckduckgo\u0027 (default, sin clave) o \u0027serpapi\u0027 (Google, requiere clave).",
      "enum": [
        "duckduckgo",
        "serpapi"
      ],
      "type": "string"
    },
    "max_results": {
      "description": "N\u00famero m\u00e1ximo de resultados (default 10, max 50).",
      "type": "integer"
    },
    "query": {
      "description": "Dork completo a ejecutar (e.g. \u0027site:example.com filetype:pdf\u0027).",
      "type": "string"
    },
    "serpapi_key": {
      "description": "Clave SerpAPI directa (alternativa a serpapi_key_binding).",
      "type": "string"
    },
    "serpapi_key_binding": {
      "description": "Nombre del binding de sesi\u00f3n que contiene la clave SerpAPI.",
      "type": "string"
    },
    "timeout_sec": {
      "description": "Timeout en segundos (default 15).",
      "type": "number"
    }
  },
  "required": [
    "query"
  ],
  "type": "object"
}

ghack_filetype

Busca archivos de un tipo específico expuestos públicamente en un dominio usando site:+filetype:. Útil para detectar documentos, configs o backups indexados.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "domain": {
      "description": "Dominio objetivo (e.g. \u0027example.com\u0027).",
      "type": "string"
    },
    "engine": {
      "description": "Motor de b\u00fasqueda (default \u0027duckduckgo\u0027).",
      "enum": [
        "duckduckgo",
        "serpapi"
      ],
      "type": "string"
    },
    "filetype": {
      "description": "Extensi\u00f3n de archivo a buscar.",
      "enum": [
        "bak",
        "cfg",
        "conf",
        "csv",
        "doc",
        "docx",
        "env",
        "gz",
        "ini",
        "json",
        "key",
        "log",
        "pdf",
        "pem",
        "ppk",
        "sql",
        "tar",
        "txt",
        "xls",
        "xlsx",
        "xml",
        "zip"
      ],
      "type": "string"
    },
    "max_results": {
      "description": "M\u00e1ximo de resultados (default 20).",
      "type": "integer"
    },
    "serpapi_key": {
      "description": "Clave SerpAPI directa.",
      "type": "string"
    },
    "serpapi_key_binding": {
      "description": "Binding de clave SerpAPI.",
      "type": "string"
    },
    "timeout_sec": {
      "description": "Timeout en segundos (default 15).",
      "type": "number"
    }
  },
  "required": [
    "domain",
    "filetype"
  ],
  "type": "object"
}

ghack_preset

Ejecuta dorks predefinidos para descubrimiento de exposición en un dominio. Presets: admin_panels, login_pages, exposed_configs, open_directories, backup_files, error_pages, database_files, api_keys.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "domain": {
      "description": "Dominio objetivo (e.g. \u0027example.com\u0027).",
      "type": "string"
    },
    "engine": {
      "description": "Motor de b\u00fasqueda (default \u0027duckduckgo\u0027).",
      "enum": [
        "duckduckgo",
        "serpapi"
      ],
      "type": "string"
    },
    "max_results": {
      "description": "M\u00e1ximo de resultados (default 20).",
      "type": "integer"
    },
    "preset": {
      "description": "Categor\u00eda de dork predefinida a ejecutar.",
      "enum": [
        "admin_panels",
        "api_keys",
        "backup_files",
        "database_files",
        "error_pages",
        "exposed_configs",
        "login_pages",
        "open_directories"
      ],
      "type": "string"
    },
    "serpapi_key": {
      "description": "Clave SerpAPI directa.",
      "type": "string"
    },
    "serpapi_key_binding": {
      "description": "Binding de clave SerpAPI.",
      "type": "string"
    },
    "timeout_sec": {
      "description": "Timeout en segundos (default 15).",
      "type": "number"
    }
  },
  "required": [
    "domain",
    "preset"
  ],
  "type": "object"
}

ghack_site_index

Enumera páginas de un dominio indexadas en buscadores usando el operador site:. Opcionalmente filtra por ruta. Útil para mapear la superficie pública de un sitio.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "domain": {
      "description": "Dominio objetivo (e.g. \u0027example.com\u0027).",
      "type": "string"
    },
    "engine": {
      "description": "Motor de b\u00fasqueda (default \u0027duckduckgo\u0027).",
      "enum": [
        "duckduckgo",
        "serpapi"
      ],
      "type": "string"
    },
    "max_results": {
      "description": "M\u00e1ximo de resultados (default 20).",
      "type": "integer"
    },
    "path_filter": {
      "description": "Ruta opcional (e.g. \u0027/admin\u0027) para acotar la b\u00fasqueda.",
      "type": "string"
    },
    "serpapi_key": {
      "description": "Clave SerpAPI directa.",
      "type": "string"
    },
    "serpapi_key_binding": {
      "description": "Binding de clave SerpAPI.",
      "type": "string"
    },
    "timeout_sec": {
      "description": "Timeout en segundos (default 15).",
      "type": "number"
    }
  },
  "required": [
    "domain"
  ],
  "type": "object"
}

Planning Extensión

Planificación

Generación de planes de alto nivel y delegación explícita.

Herramientas activas 0/1

tune Configurar variables en Ajustes

Entorno / secretos

AGENT_REGISTRY_INCLUDE_PLAN Flag Falta en BD

Default: 1

Incluir snap_plan_agent en el registry (CLI/web).
AGENT_REGISTRY_INCLUDE_DELEGATE Flag Falta en BD

Default: 1

Incluir delegate_subagent cuando hay proveedor LLM.
AGENT_DELEGATE_MAX_DEPTH Config Falta en BD

Default: 2

Profundidad máxima de delegación anidada.

snap_plan_agent

Generate a high-level plan for a given goal (Spanish); uses an LLM when available in the app.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "context": {
      "description": "Contexto adicional opcional que el modelo deber\u00eda tener en cuenta.",
      "type": "string"
    },
    "goal": {
      "description": "Objetivo o tarea a planificar (texto libre).",
      "type": "string"
    }
  },
  "required": [
    "goal"
  ],
  "type": "object"
}

Shell Extensión

Bash (snap)

Herramientas de estilo Bash del espejo snapshot.

Herramientas activas 0/5

snap_bash_command_helpers

Parse a Bash command into basic components (base command, args, pipes, redirections).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "command": {
      "description": "Bash command line to analyse.",
      "type": "string"
    }
  },
  "required": [
    "command"
  ],
  "type": "object"
}

snap_bash_permissions

Classify a Bash command in terms of mode (read/write) and destructiveness.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "command": {
      "description": "Bash command line to classify.",
      "type": "string"
    }
  },
  "required": [
    "command"
  ],
  "type": "object"
}

snap_bash_result_message

Build a human-friendly summary object for a previously executed Bash command result.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "command": {
      "description": "Executed Bash command.",
      "type": "string"
    },
    "cwd": {
      "description": "Working directory used for the command.",
      "type": "string"
    },
    "result": {
      "description": "Result object returned by snap_bash_tool (returncode, stdout, stderr).",
      "type": "object"
    }
  },
  "required": [
    "command",
    "result"
  ],
  "type": "object"
}

snap_bash_security

Provide a coarse security/risk assessment for a Bash command.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "command": {
      "description": "Bash command line to assess.",
      "type": "string"
    }
  },
  "required": [
    "command"
  ],
  "type": "object"
}

snap_bash_tool

Execute a Bash command via run_terminal with safety classification and optional destructive confirmation.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "allow_destructive": {
      "description": "Allow execution of commands classified as destructive (default false).",
      "type": "boolean"
    },
    "command": {
      "description": "Bash command line to execute.",
      "type": "string"
    },
    "cwd": {
      "description": "Optional working directory (relative to workspace root).",
      "type": "string"
    },
    "label": {
      "description": "Optional human-friendly label or note about this command.",
      "type": "string"
    },
    "timeout_sec": {
      "description": "Optional timeout in seconds for command execution.",
      "type": "number"
    }
  },
  "required": [
    "command"
  ],
  "type": "object"
}

Shell Extensión

PowerShell (snap)

Ejecución y validaciones tipo PowerShell del espejo snapshot.

Herramientas activas 0/4

snap_powershell_destructive_warning

Analyse a PowerShell command and return a destructive-command warning object if applicable.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "command": {
      "description": "PowerShell command line to analyse.",
      "type": "string"
    }
  },
  "required": [
    "command"
  ],
  "type": "object"
}

snap_powershell_path_validation

Perform a coarse validation of paths used in a PowerShell command (heuristic only).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "command": {
      "description": "PowerShell command line to analyse for path risks.",
      "type": "string"
    }
  },
  "required": [
    "command"
  ],
  "type": "object"
}

snap_powershell_readonly_validation

Validate whether a PowerShell command appears to be read-only.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "command": {
      "description": "PowerShell command line to validate.",
      "type": "string"
    }
  },
  "required": [
    "command"
  ],
  "type": "object"
}

snap_powershell_tool

Execute a PowerShell command via run_terminal with safety classification and optional destructive confirmation.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "allow_destructive": {
      "description": "Allow execution of commands classified as destructive (default false).",
      "type": "boolean"
    },
    "command": {
      "description": "PowerShell command line to execute.",
      "type": "string"
    },
    "cwd": {
      "description": "Optional working directory (relative to workspace root).",
      "type": "string"
    },
    "label": {
      "description": "Optional human-friendly label or note about this command.",
      "type": "string"
    },
    "timeout_sec": {
      "description": "Optional timeout in seconds for command execution.",
      "type": "number"
    }
  },
  "required": [
    "command"
  ],
  "type": "object"
}

Meta Extensión

Utilidades de agente

Memoria, display y utilidades internas del agente (snap_*).

Herramientas activas 0/7

snap_agent_color_manager

Resuelve colores deterministas para IDs de agentes.

Esquema

{
  "properties": {
    "ids": {
      "items": {
        "type": "string"
      },
      "type": "array"
    }
  },
  "required": [
    "ids"
  ],
  "type": "object"
}

snap_agent_display

Devuelve metainformación de presentación para un agente. Si `id` se omite, usa el agente por defecto de esta ejecución Coworker.

Esquema

{
  "properties": {
    "id": {
      "type": "string"
    }
  },
  "required": [],
  "type": "object"
}

snap_agent_memory

Lee o añade entradas de memoria asociadas a un agente (alcance por workspace registrado cuando aplica; filas sin workspace son globales legacy). Si `id` se omite, usa el agente por defecto de esta ejecución.

Esquema

{
  "additionalProperties": true,
  "properties": {
    "action": {
      "enum": [
        "get",
        "append"
      ],
      "type": "string"
    },
    "entry": {
      "additionalProperties": true,
      "properties": {
        "content": {
          "type": "string"
        },
        "kind": {
          "type": "string"
        },
        "workspace_id": {
          "type": "string"
        }
      },
      "required": [
        "content"
      ],
      "type": "object"
    },
    "id": {
      "type": "string"
    }
  },
  "required": [
    "action"
  ],
  "type": "object"
}

snap_agent_memory_snapshot

Devuelve una vista comprimida de la memoria visible en el alcance actual del workspace. `id` opcional si hay agente por defecto.

Esquema

{
  "properties": {
    "id": {
      "type": "string"
    },
    "max_entries": {
      "type": "integer"
    }
  },
  "required": [],
  "type": "object"
}

snap_agent_tool

Gestiona agentes lógicos (crear, actualizar, obtener, listar).

Esquema

{
  "additionalProperties": true,
  "properties": {
    "action": {
      "enum": [
        "list",
        "get",
        "create_or_update"
      ],
      "type": "string"
    },
    "definition": {
      "additionalProperties": true,
      "properties": {
        "color": {
          "type": "string"
        },
        "description": {
          "type": "string"
        },
        "display_name": {
          "type": "string"
        },
        "skills": {
          "items": {
            "type": "string"
          },
          "type": "array"
        },
        "system_prompt": {
          "type": "string"
        }
      },
      "type": "object"
    },
    "id": {
      "type": "string"
    }
  },
  "required": [
    "action"
  ],
  "type": "object"
}

snap_agent_tool_utils

Utilidades varias para trabajar con agentes.

Esquema

{
  "properties": {
    "action": {
      "enum": [
        "list_ids",
        "exists"
      ],
      "type": "string"
    },
    "id": {
      "type": "string"
    }
  },
  "required": [
    "action"
  ],
  "type": "object"
}

snap_list_tool_bindings

List env var and secret **names** configured for this agent run (Coworker Settings → Env vars y secrets, loaded from the database into the session). Secret **values** are never returned. Use this when the user asks for SSH or similar without naming keys: then call `snap_paramiko_ssh` with `username_binding` / `password_binding` set to the matching key names. Set `include_env_values` to true only if you need non-secret values (e.g. SSH host in env).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "include_env_values": {
      "description": "If true, include `value` for each env entry. Default false (keys only for env). Secret values are never included regardless of this flag.",
      "type": "boolean"
    }
  },
  "type": "object"
}

Archive Extensión

Archivo de skills

Inspección del subsistema de skills en disco (no confundir con skills del perfil de agente).

Herramientas activas 0/1

snap_skill_tool

Inspect the archived skills subsystem and return summary or sample files.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "action": {
      "description": "Operation to perform: \"summary\" or \"list_samples\". Default: \"summary\".",
      "type": "string"
    },
    "limit": {
      "description": "Max number of sample files to return when action=list_samples.",
      "type": "integer"
    }
  },
  "required": [],
  "type": "object"
}

Stub Extensión

Snapshot (stubs)

Espejos snapshot_* sin implementación completa o metadatos.

Herramientas activas 0/44

tune Configurar variables en Ajustes

Entorno / secretos

ENABLE_SNAPSHOT_TOOLS Flag Falta en BD

Default: 1

0/false para no registrar espejos snapshot_*.
SNAPSHOT_TOOLS_LIMIT Config Falta en BD

Límite opcional de cuántos stubs snapshot se registran.
SNAPSHOT_TOOLS_PREFIX Config Falta en BD

Default: snapshot_

Prefijo de nombre de las herramientas snapshot.
SNAPSHOT_TOOLS_FILTER Config Falta en BD

Filtro de texto opcional sobre nombres de snapshot.

Security Extensión

Criptografía y encoding

Hash (MD5/SHA/BLAKE2), Base64, HMAC, inspección JWT y generación de tokens seguros. Sin deps externas.

Herramientas activas 0/5

crypto_base64

Codifica o decodifica texto en Base64 o Base64-URL. Útil para payloads de APIs, certificados, tokens y datos binarios.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "encoding": {
      "description": "Charset del texto de entrada/salida (default \u0027utf-8\u0027).",
      "type": "string"
    },
    "input_text": {
      "description": "Texto a codificar/decodificar.",
      "type": "string"
    },
    "operation": {
      "description": "\u0027encode\u0027 o \u0027decode\u0027.",
      "enum": [
        "encode",
        "decode"
      ],
      "type": "string"
    },
    "url_safe": {
      "description": "Usar Base64-URL (- y _ en lugar de + y /, default false).",
      "type": "boolean"
    }
  },
  "required": [
    "operation",
    "input_text"
  ],
  "type": "object"
}

crypto_hash

Calcula el hash criptográfico de un texto o archivo. Algoritmos: md5, sha1, sha256 (default), sha512, sha3_256, blake2b. Salida en hexadecimal (default) o base64.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "algorithm": {
      "description": "Algoritmo de hash (default \u0027sha256\u0027).",
      "enum": [
        "blake2b",
        "md5",
        "sha1",
        "sha256",
        "sha3_256",
        "sha512"
      ],
      "type": "string"
    },
    "encoding": {
      "description": "Formato de salida: \u0027hex\u0027 (default) o \u0027base64\u0027.",
      "enum": [
        "hex",
        "base64"
      ],
      "type": "string"
    },
    "file_path": {
      "description": "Ruta de archivo a hashear (alternativa a input_text).",
      "type": "string"
    },
    "input_text": {
      "description": "Texto a hashear (alternativa a file_path).",
      "type": "string"
    }
  },
  "type": "object"
}

crypto_hmac

Genera un HMAC (Hash-based Message Authentication Code) para verificar la integridad y autenticidad de un mensaje. Algoritmos: sha256 (default), sha512.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "algorithm": {
      "description": "Algoritmo HMAC (default \u0027sha256\u0027).",
      "enum": [
        "sha256",
        "sha512"
      ],
      "type": "string"
    },
    "encoding": {
      "description": "Formato de salida (default \u0027hex\u0027).",
      "enum": [
        "hex",
        "base64"
      ],
      "type": "string"
    },
    "key": {
      "description": "Clave secreta HMAC (alternativa a key_binding).",
      "type": "string"
    },
    "key_binding": {
      "description": "Nombre del binding de sesi\u00f3n con la clave.",
      "type": "string"
    },
    "message": {
      "description": "Mensaje a autenticar.",
      "type": "string"
    }
  },
  "required": [
    "message"
  ],
  "type": "object"
}

crypto_jwt_decode

Decodifica el header y payload de un JWT sin verificar la firma. Útil para inspeccionar tokens en debug/auditoría. AVISO: no confirma que el token sea auténtico.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "token": {
      "description": "JWT en formato \u0027header.payload.signature\u0027.",
      "type": "string"
    }
  },
  "required": [
    "token"
  ],
  "type": "object"
}

crypto_random

Genera tokens o contraseñas aleatorias criptográficamente seguros. Charsets: hex (default), base64, alphanumeric, digits, printable, bytes_b64.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "charset": {
      "description": "Conjunto de caracteres (default \u0027hex\u0027).",
      "enum": [
        "alphanumeric",
        "base64",
        "digits",
        "hex",
        "printable",
        "bytes_b64"
      ],
      "type": "string"
    },
    "count": {
      "description": "Cu\u00e1ntos tokens generar (default 1, max 100).",
      "type": "integer"
    },
    "length": {
      "description": "Longitud del token en caracteres (default 32, max 4096).",
      "type": "integer"
    }
  },
  "type": "object"
}

Documents Extensión

PDF

Metadatos, extracción de texto, división y fusión de PDFs (pypdf).

Herramientas activas 0/4

tune Configurar variables en Ajustes

Entorno / secretos

pypdf Config Falta en BD

Requiere pypdf >= 5.0 (ya incluido en pyproject.toml). Sin env vars adicionales.

pdf_merge

Fusiona una lista de PDFs en un único archivo de salida.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "input_paths": {
      "description": "Lista de rutas PDF a fusionar (en orden).",
      "items": {
        "type": "string"
      },
      "type": "array"
    },
    "output_path": {
      "description": "Ruta del PDF resultante.",
      "type": "string"
    }
  },
  "required": [
    "input_paths",
    "output_path"
  ],
  "type": "object"
}

pdf_metadata

Lee los metadatos de un archivo PDF: título, autor, asunto, número de páginas, estado de cifrado y versión PDF.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "path": {
      "description": "Ruta al archivo PDF.",
      "type": "string"
    }
  },
  "required": [
    "path"
  ],
  "type": "object"
}

pdf_split

Divide un PDF en archivos de página individual (o rangos seleccionados). Los archivos se guardan en output_dir con nombre 'stem_page_NNNN.pdf'.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "output_dir": {
      "description": "Directorio donde guardar las p\u00e1ginas.",
      "type": "string"
    },
    "pages": {
      "description": "P\u00e1ginas a extraer, e.g. \u00271-3,5\u0027 (default: todas).",
      "type": "string"
    },
    "path": {
      "description": "Ruta al PDF de entrada.",
      "type": "string"
    }
  },
  "required": [
    "path",
    "output_dir"
  ],
  "type": "object"
}

pdf_text_extract

Extrae el texto de un PDF. Permite seleccionar páginas con un spec tipo '1-3,5,7' (default: todas). Límite de caracteres configurable.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "max_chars": {
      "description": "L\u00edmite de caracteres en la salida (default 50 000).",
      "type": "integer"
    },
    "pages": {
      "description": "P\u00e1ginas a extraer, e.g. \u00271-3,5\u0027 (default: todas).",
      "type": "string"
    },
    "path": {
      "description": "Ruta al archivo PDF.",
      "type": "string"
    }
  },
  "required": [
    "path"
  ],
  "type": "object"
}

Network Extensión

Cliente HTTP

Llamadas HTTP/HTTPS genéricas y atajo para APIs JSON con auth Bearer/API-key. Sin deps externas.

Herramientas activas 0/2

http_json_api

Atajo para llamadas a APIs JSON. Serializa/deserializa automáticamente. Soporta Bearer token y API-key vía binding de sesión. Permite construir la URL con base_url + path.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "api_key": {
      "description": "API key directa.",
      "type": "string"
    },
    "api_key_binding": {
      "description": "Binding de sesi\u00f3n para la API key.",
      "type": "string"
    },
    "api_key_header": {
      "description": "Nombre del header de API key (default \u0027X-API-Key\u0027).",
      "type": "string"
    },
    "base_url": {
      "description": "URL base (e.g. \u0027https://api.example.com\u0027).",
      "type": "string"
    },
    "bearer_token": {
      "description": "Bearer token directo.",
      "type": "string"
    },
    "bearer_token_binding": {
      "description": "Binding de sesi\u00f3n para el Bearer token.",
      "type": "string"
    },
    "headers": {
      "description": "Headers adicionales.",
      "type": "object"
    },
    "json_body": {
      "description": "Payload JSON a enviar.",
      "type": "object"
    },
    "method": {
      "description": "M\u00e9todo HTTP (default \u0027GET\u0027).",
      "enum": [
        "DELETE",
        "GET",
        "HEAD",
        "OPTIONS",
        "PATCH",
        "POST",
        "PUT"
      ],
      "type": "string"
    },
    "path": {
      "description": "Path relativo (e.g. \u0027/v1/users\u0027).",
      "type": "string"
    },
    "timeout_sec": {
      "description": "Timeout en segundos (default 30).",
      "type": "number"
    },
    "url": {
      "description": "URL completa (alternativa a base_url+path).",
      "type": "string"
    },
    "verify_ssl": {
      "description": "Verificar SSL (default false).",
      "type": "boolean"
    }
  },
  "type": "object"
}

http_request

Realiza una llamada HTTP genérica (GET/POST/PUT/DELETE/PATCH/HEAD). Soporta headers, body, JSON, autenticación Bearer/Basic/API-key (con binding de sesión), control de SSL y redirects.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "api_key_header": {
      "description": "Nombre del header para API key (default \u0027X-API-Key\u0027).",
      "type": "string"
    },
    "auth_binding": {
      "description": "Nombre de binding de sesi\u00f3n para el valor de auth.",
      "type": "string"
    },
    "auth_type": {
      "description": "Tipo de autenticaci\u00f3n.",
      "enum": [
        "bearer",
        "basic",
        "api_key",
        "none"
      ],
      "type": "string"
    },
    "auth_value": {
      "description": "Valor de autenticaci\u00f3n directo (bearer token, user:pass, api key).",
      "type": "string"
    },
    "body": {
      "description": "Cuerpo de la petici\u00f3n como texto.",
      "type": "string"
    },
    "follow_redirects": {
      "description": "Seguir redirects (default true).",
      "type": "boolean"
    },
    "headers": {
      "description": "Headers adicionales como objeto clave-valor.",
      "type": "object"
    },
    "json_body": {
      "description": "Cuerpo JSON (auto-serializado; a\u00f1ade Content-Type: application/json).",
      "type": "object"
    },
    "max_response_bytes": {
      "description": "Bytes m\u00e1ximos de respuesta (default 512 KB).",
      "type": "integer"
    },
    "method": {
      "description": "M\u00e9todo HTTP (default \u0027GET\u0027).",
      "enum": [
        "DELETE",
        "GET",
        "HEAD",
        "OPTIONS",
        "PATCH",
        "POST",
        "PUT"
      ],
      "type": "string"
    },
    "timeout_sec": {
      "description": "Timeout en segundos (default 30).",
      "type": "number"
    },
    "url": {
      "description": "URL completa (http:// o https://).",
      "type": "string"
    },
    "verify_ssl": {
      "description": "Verificar certificado SSL (default false).",
      "type": "boolean"
    }
  },
  "required": [
    "url"
  ],
  "type": "object"
}

Data Extensión

SQL / SQLite

Consultas SELECT, DML, listado de tablas y descripción de esquema sobre archivos SQLite.

Herramientas activas 0/3

sql_describe_table

Describe el esquema de una tabla SQLite: columnas, tipos, índices, claves foráneas y número de filas.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "db_binding": {
      "description": "Binding de sesi\u00f3n con la ruta a la BD.",
      "type": "string"
    },
    "db_path": {
      "description": "Ruta al archivo .db/.sqlite.",
      "type": "string"
    },
    "table_name": {
      "description": "Nombre exacto de la tabla.",
      "type": "string"
    }
  },
  "required": [
    "table_name"
  ],
  "type": "object"
}

sql_list_tables

Lista las tablas (y vistas) de un archivo SQLite.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "db_binding": {
      "description": "Binding de sesi\u00f3n con la ruta a la BD.",
      "type": "string"
    },
    "db_path": {
      "description": "Ruta al archivo .db/.sqlite.",
      "type": "string"
    },
    "include_views": {
      "description": "Incluir vistas adem\u00e1s de tablas (default true).",
      "type": "boolean"
    }
  },
  "type": "object"
}

sql_query

Ejecuta una consulta SQL sobre un archivo SQLite. Las consultas SELECT son solo lectura por defecto; pasa allow_write=true para INSERT/UPDATE/DELETE. Soporta placeholders posicionales (?).

Esquema

{
  "additionalProperties": false,
  "properties": {
    "allow_write": {
      "description": "Permitir INSERT/UPDATE/DELETE/DDL (default false).",
      "type": "boolean"
    },
    "db_binding": {
      "description": "Binding de sesi\u00f3n con la ruta a la BD (alternativa a db_path).",
      "type": "string"
    },
    "db_path": {
      "description": "Ruta al archivo .db/.sqlite.",
      "type": "string"
    },
    "max_rows": {
      "description": "M\u00e1ximo de filas a devolver (default 500, max 10 000).",
      "type": "integer"
    },
    "params": {
      "description": "Par\u00e1metros posicionales para placeholders ? (opcional).",
      "items": {},
      "type": "array"
    },
    "sql": {
      "description": "Sentencia SQL a ejecutar.",
      "type": "string"
    }
  },
  "required": [
    "sql"
  ],
  "type": "object"
}

Alerts Extensión

Notificaciones

Envío de correo (SMTP/STARTTLS), webhooks HTTP genéricos y mensajes Slack vía Incoming Webhook.

Herramientas activas 0/3

tune Configurar variables en Ajustes

Entorno / secretos

SMTP_HOST · SMTP_USERNAME · SMTP_PASSWORD Secreto Falta en BD

Cuándo: notify_email

Credenciales SMTP. Pueden pasarse directamente en la llamada o vía binding de sesión.
SLACK_WEBHOOK_URL Secreto Falta en BD

Cuándo: notify_slack

URL del Incoming Webhook de Slack. Puede pasarse vía webhook_binding.

notify_email

Envía un correo electrónico vía SMTP. Soporta STARTTLS (port 587 default) y SSL directo (port 465). Credenciales mediante binding de sesión o valor directo. El cuerpo puede ser texto plano o HTML.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "body": {
      "description": "Cuerpo del mensaje.",
      "type": "string"
    },
    "cc": {
      "description": "CC opcional.",
      "oneOf": [
        {
          "type": "string"
        },
        {
          "items": {
            "type": "string"
          },
          "type": "array"
        }
      ]
    },
    "content_type": {
      "description": "Tipo de contenido del cuerpo (default \u0027plain\u0027).",
      "enum": [
        "plain",
        "html"
      ],
      "type": "string"
    },
    "from_addr": {
      "description": "Direcci\u00f3n de remitente (default: username).",
      "type": "string"
    },
    "password": {
      "description": "Contrase\u00f1a SMTP (preferir password_binding).",
      "type": "string"
    },
    "password_binding": {
      "description": "Binding de sesi\u00f3n para la contrase\u00f1a SMTP.",
      "type": "string"
    },
    "smtp_host": {
      "description": "Servidor SMTP.",
      "type": "string"
    },
    "smtp_host_binding": {
      "description": "Binding de sesi\u00f3n para el host SMTP.",
      "type": "string"
    },
    "smtp_port": {
      "description": "Puerto SMTP (default 587).",
      "type": "integer"
    },
    "subject": {
      "description": "Asunto del correo.",
      "type": "string"
    },
    "to": {
      "description": "Destinatario/s (string o lista de emails).",
      "oneOf": [
        {
          "type": "string"
        },
        {
          "items": {
            "type": "string"
          },
          "type": "array"
        }
      ]
    },
    "use_ssl": {
      "description": "Usar SSL desde el inicio, port 465 (default false).",
      "type": "boolean"
    },
    "use_tls": {
      "description": "Usar STARTTLS (default true).",
      "type": "boolean"
    },
    "username": {
      "description": "Usuario SMTP.",
      "type": "string"
    },
    "username_binding": {
      "description": "Binding de sesi\u00f3n para el usuario SMTP.",
      "type": "string"
    }
  },
  "required": [
    "to",
    "body"
  ],
  "type": "object"
}

notify_slack

Envía un mensaje a Slack vía Incoming Webhook. Soporta canal, icono, username personalizado y campos de attachment. La URL del webhook puede estar en un binding de sesión.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "channel": {
      "description": "Canal destino (p. ej. \u0027#alerts\u0027), sobreescribe el destino del webhook.",
      "type": "string"
    },
    "color": {
      "description": "Color del attachment en hex (default \u0027#36a64f\u0027).",
      "type": "string"
    },
    "fields": {
      "description": "Campos adicionales en el attachment (title/value/short).",
      "items": {
        "properties": {
          "short": {
            "type": "boolean"
          },
          "title": {
            "type": "string"
          },
          "value": {
            "type": "string"
          }
        },
        "type": "object"
      },
      "type": "array"
    },
    "icon_emoji": {
      "description": "Emoji de icono, p. ej. \u0027:robot_face:\u0027.",
      "type": "string"
    },
    "text": {
      "description": "Texto principal del mensaje.",
      "type": "string"
    },
    "timeout_sec": {
      "description": "Timeout en segundos (default 10).",
      "type": "number"
    },
    "username": {
      "description": "Nombre del bot remitente.",
      "type": "string"
    },
    "webhook_binding": {
      "description": "Binding de sesi\u00f3n para la URL del webhook.",
      "type": "string"
    },
    "webhook_url": {
      "description": "URL del Incoming Webhook de Slack.",
      "type": "string"
    }
  },
  "required": [
    "text"
  ],
  "type": "object"
}

notify_webhook

Envía una notificación POST a un webhook HTTP genérico (n8n, Zapier, Make, webhook.site, etc.). Acepta payload JSON o form-data. La URL puede estar en un binding de sesión.

Esquema

{
  "additionalProperties": false,
  "properties": {
    "auth_binding": {
      "description": "Binding de sesi\u00f3n para el header Authorization completo.",
      "type": "string"
    },
    "form_data": {
      "description": "Datos como form-urlencoded (alternativa a payload).",
      "type": "object"
    },
    "headers": {
      "description": "Headers adicionales.",
      "type": "object"
    },
    "payload": {
      "description": "Payload JSON a enviar (alternativa a form_data).",
      "type": "object"
    },
    "timeout_sec": {
      "description": "Timeout en segundos (default 15).",
      "type": "number"
    },
    "url": {
      "description": "URL del webhook (http/https).",
      "type": "string"
    },
    "url_binding": {
      "description": "Binding de sesi\u00f3n para la URL del webhook.",
      "type": "string"
    }
  },
  "type": "object"
}