What it does
The MCP server exposes the AgentFactory fleet orchestration capabilities to any MCP-aware client. It acts as a bridge between external tools (like Claude Code) and the Donmai fleet runtime, allowing clients to spawn, monitor, and manage a pool of coding agents working across development, QA, and acceptance pipelines. The server handles agent lifecycle management, work queue dispatch, and status reporting so MCP clients can orchestrate multi-agent workflows without needing direct access to the underlying orchestrator.
Who it's for
Teams running a Donmai or AgentFactory fleet in production who need external tools to programmatically manage their agents. Developers building MCP integrations for multi-agent workflows, and organizations automating issue resolution or code generation through agent orchestration.
Common use cases
- Spawn and manage coding agents from Claude Code or other MCP clients without direct access to the fleet
- Query agent fleet status, workload, and progress from external monitoring tools
- Submit issues or pull requests to the orchestrator queue for agent processing
- Retrieve agent results and integrate them back into CI/CD pipelines or chat interfaces
- Automate multi-agent workflows for large-scale code generation or issue resolution
Setup pitfalls
- Requires Redis for work queue and session storage — ensure Redis is running and accessible before starting the server
LINEAR_ACCESS_TOKENenvironment variable required if integrating with Linear issue tracker- Server reads and writes to the filesystem for agent worktrees and logs — sandbox permissions carefully
- Classified as high-risk due to network calls and filesystem access — audit all connected infrastructure before deployment