CodyMaster + Goose Integration Guide

Goose is an open-source AI agent with 55,000+ users. CodyMaster adds persistent memory, skill orchestration, and token management to any Goose session.

Prerequisites

• Goose installed (pip install goose-ai or see Goose docs)
• CodyMaster installed (npm install -g codymaster)
• A project directory to work in

Installation

1. Install CodyMaster

npm install -g codymaster

2. Get the Goose configuration snippet

cd /path/to/your/project
cm mcp-serve --print-config

This outputs:

{
  "id": "codymaster",
  "name": "CodyMaster Intelligence Layer",
  "type": "stdio",
  "cmd": "npx",
  "args": ["codymaster", "mcp-serve", "--project", "/path/to/your/project"]
}

3. Add to Goose config

Add the extension to your ~/.config/goose/config.yaml:

extensions:
  - id: codymaster
    name: CodyMaster Intelligence Layer
    type: stdio
    cmd: npx
    args:
      - codymaster
      - mcp-serve
      - --project
      - /path/to/your/project
    enabled: true

4. Restart Goose

Start a new Goose session. CodyMaster tools will appear in the available tools list.

Testing the Integration

In your Goose session, try:

Use cm_query to search for any existing project learnings.

Or use the natural language interface:

Remember that we use pnpm, not npm, for this project.

What did we learn about the authentication module?

Available Tools

Tool	Description
cm_query	FTS5 search across learnings + decisions
cm_memory_write	Save a new learning/knowledge
cm_natural	Natural language interface ("remember X", "forget Y", "what did we learn about Z")
cm_budget_check	Check token budget before loading large contexts
cm_bus_read	Read current skill pipeline state
cm_plan	Get sprint context + next skill hint
cm_deploy	Deploy workflow hints

Troubleshooting

cm mcp-serve not found: Run npm run build in the CodyMaster project directory first.

No tools appearing in Goose: Check that enabled: true is set in your Goose config and restart Goose.

Memory not persisting: Ensure --project points to the same directory in each session. Memory is stored in .cm/context.db.