Documentation
donmai is the OSS CLI and daemon for AI agent fleets. Start here.
donmai (どんまい) is the open-source CLI and daemon for running AI agent
fleets: one Go binary that ships a persistent daemon, a dispatch loop,
plugin-delivered MCP tools (embedded in the binary), and a typed workflow
YAML grammar from the @donmai/core TypeScript package. MIT licensed. The same runtime powers Rensei in production.
Four commands to a running daemon:
brew install RenseiAI/homebrew-tap/donmai
donmai daemon setup
donmai daemon install
donmai daemon statusdonmai is 0.x preview; APIs may change between minor releases. Other install paths (go install, release tarball, building from source) are on the install page.
Architecture overview
The stack has two layers. The OSS layer is the donmai binary (Go) with its
daemon and dispatch loop, plus @donmai/core (TypeScript) for the workflow
YAML grammar and contracts. MCP tool plugins (Linear and code intelligence)
ship embedded in the binary. The durable executor that runs workflow
definitions at scale, the concrete node library (about 60 nodes), the visual
editor, and cloud execution are the Rensei platform, closed source. The
contract between the two is documented here, exactly.
Sections
getting started
install the binary, run the setup wizard, dispatch a first agent session.
configuration
the ~/.donmai/daemon.yaml key reference and every environment variable.
providers
the provider plugin contract and the capability matrix.
plugins
ToolPlugin packages and the MCP tools the binary ships.
workflow nodes
ActionDefinition, TriggerDefinition, and ConditionDefinition contracts, and the workflow YAML grammar.
Source
Source is on GitHub at github.com/RenseiAI/donmai. Issues and pull requests welcome.