███╗ ███╗ ██████╗ █████╗ ██████╗ ███████╗███╗ ██╗████████╗ ████╗ ████║██╔═══██╗ ██╔══██╗██╔════╝ ██╔════╝████╗ ██║╚══██╔══╝ ██╔████╔██║██║ ██║█████╗███████║██║ ███╗█████╗ ██╔██╗ ██║ ██║ ██║╚██╔╝██║██║▄▄ ██║╚════╝██╔══██║██║ ██║██╔══╝ ██║╚██╗██║ ██║ ██║ ╚═╝ ██║╚██████╔╝ ██║ ██║╚██████╔╝███████╗██║ ╚████║ ██║ ╚═╝ ╚═╝ ╚══▀▀═╝ ╚═╝ ╚═╝ ╚═════╝ ╚══════╝╚═╝ ╚═══╝ ╚═╝
mq-agent
Terminal-native AI agent orchestrator for the mq ecosystem.
What it is
A controlled terminal workflow for AI agent work — explicit planning, tool routing, verification, persistent memory and safety gates. Every action is traceable. The model never runs unsupervised.
Planner
Decomposes goals into ordered steps via structured JSON output.
MQ_AGENT_MODELExecutor
Routes steps through a typed tool registry with dry-run support.
Verifier
Checks each step result before advancing.
MQ_AGENT_VERIFIER_MODELSafety
Four modes: read-only → suggest → execute → dangerous.
Memory
Persistent JSON store — carries context across sessions.
MCP Bridge
Routes tool calls to mq-mcp with safety classification.
Quick start
$ uv pip install -e ".[dev,signal]" $ mq-agent doctor # check environment $ mq-agent score . # README + publish score $ mq-agent dashboard # operator snapshot $ mq-agent signal . # repo-signal assessment $ mq-agent audit . # read-only repo audit
Commands
No API key needed
mq-agent doctor # environment check mq-agent score . # README + publish score mq-agent repo-summary . # repo overview mq-agent tools # list registered tools mq-agent dashboard # operator snapshot mq-agent stack loop # controlled loop plan mq-agent tui # refreshable Textual dashboard
AI-powered (needs OPENAI_API_KEY)
mq-agent audit . # read-only repo audit mq-agent docs-audit . # audit README, CHANGELOG and /docs mq-agent signal . # repo-signal + AI improvement plan mq-agent plan "goal" # decompose a goal into steps mq-agent release-check # validate release readiness mq-agent fix-ci # diagnose CI failures
MCP local tools
mq-agent mcp status # check mq-mcp reachability mq-agent mcp tools # list tools with safety classes mq-agent run-tool <tool> # run tool through safety gates