AgentRQ
A self-hosted task manager for human and agent collaboration through MCP
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- The agent creates and changes tasks in the workspace through MCP
- Requires a deployed network service with access tokens
Install
Manual install
{
"mcpServers": {
"agentrq-WORKSPACE_ID": {
"type": "http",
"url": "YOUR_MCP_URL"
}
}
}The URL with token comes from the workspace setup modal; place .mcp.json in the project folder.
This is third-party code. Review the repository files before installing.
What it does
AgentRQ is a shared workspace where a human and AI agents run tasks together. Through MCP the agent sees the workspace state: it pulls assigned tasks, updates statuses, requests permission for important actions and replies to the human, all synced across devices. It has a task board and kanban, scheduling and recurring task launches, named events that spawn new tasks, and workflows as a node graph. It connects to Claude through a project .mcp.json with the workspace URL and token. The product is open-source and self-hosted, including via Docker.
Who it is for. For teams and developers who want to assign tasks to agents and track their work in one place.
Good fit when
- You want to give the agent tasks and see their status on a board
- You want a human in the loop for the agent's important actions
- You need schedules, recurring tasks and events that trigger work
Not a fit when
- You do not need a separate task manager on top of the agent
- You cannot deploy and maintain a self-hosted service
Example request
Pull my assigned tasks from AgentRQ and mark the finished one as doneLimitations
It is a self-hosted platform you deploy and keep running, with a Docker image and a separate guide for production. The agent connects over MCP through a URL with the workspace token. Setup assumes working with the project config and permissions.
How to disable. Remove the agentrq server from the project's .mcp.json and stop the deployed service.
MCP
- Transport
- http
- Authentication
- API key
Security check
- The agent creates and changes tasks in the workspace through MCP
- Requires a deployed network service with access tokens
README in short
The README describes AgentRQ as an MCP-based human and agent collaboration platform: the agent sees the workspace state, pulls tasks, changes statuses, requests permissions and replies. It has a board and kanban, scheduling, recurring tasks, named events and node-graph workflows. Connecting to Claude goes through .mcp.json with a URL and token, and pre-approving the needed tools is suggested for important actions. The project is self-hosted, including via Docker. Apache-2.0 licensed.
FAQ
How does the agent see tasks?
Through MCP the agent reads the workspace state, pulls tasks, updates statuses and replies to the human.
Can I self-host it?
Yes, the project is open-source and self-hosted, with a ready Docker image for production.
Related
A self-hosted knowledge base with block-level references and a built-in MCP server for connecting AI agents to your notes
A CLI for every Google Workspace API with JSON output and agent skills: Drive, Gmail, Calendar, Sheets and more
Local search over Markdown notes, docs and meeting transcripts: keywords, semantic search and reranking, with an MCP server
A task manager for AI-driven development: breaks a PRD into dependent tasks and guides the agent through them via MCP or CLI