-
Notifications
You must be signed in to change notification settings - Fork 0
Integrations
mq-agent is the orchestration layer for the mq ecosystem. It routes goals to the right tools, agents and services.
Role: Command surface and menu entry point.
mqlaunch is the top-level CLI that users interact with. It delegates tasks to mq-agent for orchestration. mq-agent should be callable directly or via mqlaunch.
Status: Active. See COMMAND_SURFACE.md and MQLAUNCH_INTEGRATION.md.
Role: Reasoning and local assistant layer.
mq-hal provides local AI reasoning capabilities. Future integration may let mq-agent use mq-hal as a fallback planner when OpenAI is not available, or as a local-first reasoning layer for faster, cheaper planning on simple tasks.
Status: Planned.
Role: Tool execution and local bridge.
mq-mcp exposes tools via the Model Context Protocol over HTTP. mq-agent connects to mq-mcp via mq_agent/tools/mcp_bridge.py.
When mq-mcp is running locally on :8765, mcp_call can route tool calls to it:
from mq_agent.tools.mcp_bridge import MCPBridge
bridge = MCPBridge("http://localhost:8765")
result = bridge.call_tool("your_tool", {"arg": "value"})Check availability:
mq-agent doctor # shows mq-mcp (optional) statusStatus: Active. Tool discovery, tool listing, metadata inspection and gated tool execution are implemented.
Role: Repository intelligence, scoring and release readiness.
repo-signal analyses repositories and produces structured health scores and release readiness signals. mq-agent integrates repo-signal output into:
-
mq-agent audit— health score per repo -
mq-agent release-check— release readiness score with blocking criteria -
mq-agent repo-summary— signal-enriched summary
Status: Active.
Role: Planning and verification backbone.
mq-agent uses two OpenAI models:
| Model | Role |
|---|---|
gpt-4o |
Planner — goal decomposition |
gpt-4o-mini |
Verifier — per-step result checking |
Both use structured JSON output mode. Prompts live in prompts/planner.md and prompts/verifier.md.
Requires OPENAI_API_KEY in environment.
- Semantic memory — OpenAI vector stores for cross-session repo knowledge
- Local model support — Ollama or similar for offline planning