UseAgent fleet MCP
UseAgent’s MCP server sends tasks to remote agents, collects their results and records the runs in an organization’s ledger.
- 370 ★ on the repository
- Last commit
- AGPL-3.0
Broad access
Why this level
- Remote agents can execute commands and modify connected repositories using their granted access.
Install
- Works in
Manual install
Read the CLI package guide in the UseAgent repository. Prepare a source-based Bun launch with the organization key in the environment. Test one task and inspect its result and permissions before dispatching parallel jobs.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Remote agents can execute commands and modify connected repositories using their granted access.
This is third-party code. Review the repository files before installing.
What it does
UseAgent’s MCP server lets a local agent dispatch jobs into an organization’s workspace. It can send a single task, parallel tasks or a small batch with a shared identifier. Each run exposes its result and a conversation link. The server is bundled with UseAgent’s CLI. The platform runs Claude Code, Codex and OpenCode in separate Linux environments. Result collection can request a verifier that posts its judgment in the same conversation.
Who it is for. Technical leads with a UseAgent deployment distributing repeatable jobs across several agents.
Good fit when
- Several independent repositories need the same check and consolidated results.
- A local agent should hand off a long task and retrieve the answer later.
- Completed results need a separate verification step in the same work log.
- After a control-server restart, you need the history of dispatched work.
Not a fit when
- You only need a local assistant without an organization server or remote environments.
- Code must not reach the selected model or environment provider.
- Stable APIs are mandatory and the team cannot pin and test versions.
Example request
Dispatch documentation reviews for these repositories to three agents and collect the answers with verifier judgments.MCP
- Transport
- stdio
- Authentication
- API key
Environment variables
USEAGENT_API_KEYrequired, secret- UseAgent organization API key.
USEAGENT_BASE_URL- Base URL of your UseAgent deployment.
Limitations
The CLI’s own guide still marks the package as unpublished: run it from source and check package availability separately before using bunx. Real tasks require a UseAgent organization key, a configured model and an isolated-environment provider. The platform is alpha software, so APIs and schemas may change between releases. Self-hosting requires Bun and PostgreSQL 16+ with pgvector; starting the web interface alone does not provision an agent execution environment.
How to disable. Remove the local agent’s server connection and revoke the organization key. Review already dispatched jobs separately in UseAgent.
FAQ
Must dispatch wait for the answer?
No. MCP dispatch tools return run identifiers. Retrieve each result in a separate request.
What does the verifier check?
It follows your verification criterion and posts PASS or FAIL next to the work. Specify the criterion and read its explanation.
How do I inspect the source-based CLI?
Run `bun packages/cli/src/bin.ts --help` in the repository. Help lists the commands; live runs need an organization key.
Can tasks use different models?
Yes. Tasks can specify their own engine and model. Check that the selected combination is available to your organization.
Where can I find recent tasks?
The MCP server has an organization recent-runs tool. Use a returned identifier to fetch status and the answer.
Discussion
No questions or reviews yet.
Related
MCP serversA bridge between local CLI agents: Codex, Cursor or Claude Code delegate work to Antigravity, Grok, Kimi and seven more.
An MCP server and CLI coordinate specialized agents, retain working patterns and connect tasks with project hooks.
AgentOps MCP server: the agent reads traces and spans of agent runs to find where they failed
MCP Agent Mail
mcp_agent_mail
A mail coordination layer for coding agents: agents have memorable names, inboxes and outboxes, history search and voluntary file reservations
A CLI that turns any MCP server, OpenAPI spec or GraphQL endpoint into commands at runtime, with no codegen
A mail server for agents sharing one repository: messages, threads, file reservations and a 16-screen TUI
README in short
PostgreSQL records each run with tool calls and approvals. Integration credentials stay on the control server rather than entering the agent environment.