Skip to main content
Glama
openclaw-task-mcp

openclaw-task-mcp

openclaw-task-mcp

Simple Node stdio MCP server for OpenClaw task lifecycle.

Tools:

  • start_task → sessions_spawn

  • list_tasks → sessions_list (subagent-filtered)

  • get_task_status → sessions_history (+ result when available)

Requirements

  • Node.js >= 18.19.1

  • Reachable OpenClaw Gateway

  • Gateway token with permission to invoke tools

Related MCP server: OpenClaw MCP Server

Install

git clone https://github.com/EnviableAmaretto/openclaw-task-mcp.git
cd openclaw-task-mcp
npm install
npm run hooks:install

MCP config (Claude-code style / SillyTavern-style)

Put base URL and token directly in the MCP server env block:

{
  "mcpServers": {
    "openclaw": {
      "command": "node",
      "args": ["/ABSOLUTE/PATH/TO/openclaw-task-mcp/src/index.js"],
      "env": {
        "OPENCLAW_BASE_URL": "http://127.0.0.1:18789",
        "OPENCLAW_API_TOKEN": "your-gateway-token",
        "OPENCLAW_TIMEOUT_MS": "30000"
      }
    }
  }
}

args[0] must be the absolute path where you cloned this repository on your host. Examples:

  • /home/sillytavern/SillyTavern/repositories/openclaw-task-mcp/src/index.js

  • /opt/openclaw-task-mcp/src/index.js

  • /srv/mcp/openclaw-task-mcp/src/index.js

No separate HTTP MCP endpoint is needed in this mode; the app launches this process over stdio.

Local run (manual)

OPENCLAW_BASE_URL="http://127.0.0.1:18789" \
OPENCLAW_API_TOKEN="YOUR_TOKEN" \
node src/index.js

Tests

npm test

Quality gates

  • Local pre-push hook (.githooks/pre-push) runs npm test.

  • CI (.github/workflows/ci.yml) runs npm ci && npm test on push/PR.

  • Mandatory before commit: launch MCP server (npm start) and execute at least one real E2E MCP call successfully. Tests alone are not sufficient.

Security notes

  • Keep token private; do not commit it.

  • Prefer private host networking for the OpenClaw URL.

  • Rotate token regularly.

Available Tools

3 tools
get_task_statusA

Get task status from sessions_history and include result when available

Inputs: sessionId (required), includeResult.

list_tasksC

List tasks via sessions_list, filtered to subagent by default

Inputs: limit, cursor, subagentOnly.

start_taskC

Start a subagent task via OpenClaw sessions_spawn

Inputs: title (required), prompt (required), channel, metadata, +1 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.3/5.0

    Scored across 3 tools

    Disambiguation5/5

    Each tool targets a distinct action: creating a task, listing tasks, and checking status. There is no overlap or ambiguity between purposes.

    Naming Consistency5/5

    All tool names follow a consistent verb_noun pattern: start_task, list_tasks, get_task_status. The naming is uniform and predictable.

    Tool Count5/5

    Three tools is a well-scoped set for task lifecycle management. Each tool earns its place and there is no redundancy or bloat.

    Completeness4/5

    The core task lifecycle is covered: starting, listing, and retrieving status. A minor gap is the lack of an explicit cancel/abort operation, but the primary workflow works without dead ends.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers