Built-in Tools

MindRoom includes 100+ built-in tool integrations organized by category.

File & System

Web Search & Research

Web Scraping & Crawling

AI & ML APIs

Knowledge & Research

Communication & Social

thread_tags

tag_thread adds or updates one tag on the active thread by default. untag_thread removes one tag from the active thread by default. Passing room_id for the current room does not disable that same-room thread fallback for the write operations. list_thread_tags also defaults to the active thread. To list every tagged thread in a room from inside an active thread, pass room_id without thread_id. list_thread_tags(tag=...) returns only the requested tag in both thread-specific and room-wide responses. The predefined blocked payload accepts data.blocked_by as a list of non-empty strings. The predefined waiting payload accepts data.waiting_on as one non-empty string. The predefined priority payload accepts data.level as high, medium, or low. The predefined due payload accepts data.deadline as an ISO-8601 timestamp. All predefined payloads are normalized before they are persisted, and malformed persisted tag payloads are ignored on read. This intentionally replaces the removed experimental thread_resolution tool and does not auto-read old com.mindroom.thread.resolution markers.

Project Management

Calendar & Scheduling

Data & Business

Location & Maps

DevOps & Infrastructure

Smart Home

Media & Entertainment

Memory & Storage

Custom & Config

Tool presets are config-only macros, not runtime tools. For OpenClaw workspace portability, openclaw_compat expands to shell, coding, duckduckgo, website, browser, scheduler, subagents, matrix_message, and attachments. attachments is not directly in the preset — it is implied by matrix_message via the IMPLIED_TOOLS mapping.

Claude Agent Sessions

The claude_agent tool manages long-lived Claude coding sessions on the backend. This allows iterative coding workflows in the same session (including Claude-side tool usage and subagents).

When using the OpenAI-compatible API, set X-Session-Id to keep tool sessions stable across requests. See OpenAI API Compatibility.

Add claude_agent to an agent's tools in config.yaml:

agents:
  code:
    display_name: Code Agent
    role: Coding assistant with persistent Claude sessions
    model: general
    tools:
      - claude_agent

Configure credentials via the dashboard or by writing mindroom_data/credentials/claude_agent_credentials.json:

{
  "api_key": "sk-ant-or-proxy-key",
  "model": "claude-sonnet-4-6",
  "permission_mode": "default",
  "continue_conversation": true,
  "session_ttl_minutes": 60,
  "max_sessions": 200
}

To run through an Anthropic-compatible gateway (for example LiteLLM /v1/messages):

{
  "api_key": "sk-dummy",
  "anthropic_base_url": "http://litellm.local",
  "anthropic_auth_token": "sk-dummy",
  "disable_experimental_betas": true
}

Use the gateway host root for anthropic_base_url (no /v1 suffix), because Claude clients append /v1/messages. Some Anthropic-compatible backends may reject Claude's anthropic-beta headers. Set disable_experimental_betas to true in that case.

Worker-Routed Tools

Some tools default to running in a sandboxed worker container instead of the primary agent process. This is controlled by the default_execution_target metadata field.

The following tools default to worker execution:

When a sandbox proxy backend is configured, calls to these tools are forwarded to an isolated container. The worker container receives credentials via short-lived leases and inherits the agent's execution scope.

Execution Scopes

Worker routing uses three scope levels configured via worker_scope on an agent:

Shared-Only Integrations

Some dashboard integrations are restricted to shared or unscoped execution and cannot be used by agents with user or user_agent worker scope:

google, spotify, homeassistant, gmail, google_calendar, google_sheets

If an agent with an isolating scope tries to use one of these integrations, the tool call is rejected with an error explaining the scope restriction.

Enabling Tools

Add tools to agents in config.yaml — as plain strings or single-key dicts with inline config overrides:

agents:
  assistant:
    display_name: Assistant
    role: A helpful assistant
    model: sonnet
    tools:
      - file
      - shell:                            # per-agent config override
          extra_env_passthrough: "DAWARICH_*"
      - duckduckgo
      - github

Or use the Dashboard's Agents tab to enable tools visually. See Per-Agent Tool Configuration for the full override syntax.

Environment Variables

MindRoom still supports a config-adjacent .env file for provider and bootstrap settings such as model API keys:

# AI providers
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=...
REPLICATE_API_TOKEN=r8_...

That .env file is resolved next to the active config.yaml, not from an arbitrary working directory.

Tool integrations should not rely on tool-specific env vars such as CLICKUP_API_KEY or DAYTONA_API_KEY.

Configure tools through the dashboard credentials store, persisted tool configuration, or explicit runtime overrides instead.

Execution tools do not use env vars as constructor-time configuration. shell still receives the committed runtime env as explicit execution context. Additional process env must be opted into with extra_env_passthrough. If shell execution needs extra PATH entries such as wrapper directories, configure shell_path_prepend instead of relying on host-specific defaults. When run_shell_command hits its timeout, it keeps the command running in the background and returns a handle. Use check_shell_command(handle) to poll status and kill_shell_command(handle) to stop a backgrounded process. These handles are process-local: they survive multiple requests to the same runner process, but do not survive runner restarts. python should not rely on in-process runtime env emulation. If Python code needs runtime-scoped env, run it through sandbox/worker subprocess execution instead.