cacodes-mcp-server 0.3.0

Model Context Protocol server for Cacodes (Plane fork) integration

CaCodes MCP Server

Model Context Protocol server for CaCodes (Plane fork) — work items, pages, project memory (AgentMemory RAG), and templates.

PyPI: cacodes-mcp-server
Remote: https://mcp.cacodes.com.br/http/api-key/mcp

Quick install

uvx cacodes-mcp-server install --all
uvx cacodes-mcp-server doctor
uvx cacodes-mcp-server stdio

The install orchestrator configures MCP, runs the PAT wizard (or --auth oauth), validates connectivity, installs GSD and Ponytail from local vendor snapshots, and writes memory-first Cursor rules.

Subcommands

Command	Description
install	MCP + GSD + Ponytail + rules + optional doctor
doctor	Validate PAT, workspace, MCP, memory_recall
auth	PAT wizard only (127.0.0.1:8765/callback)
stdio / http	Run MCP server

Cursor (HTTP remote, PAT)

{
  "mcpServers": {
    "cacodes": {
      "url": "https://mcp.cacodes.com.br/http/api-key/mcp",
      "type": "http",
      "headers": {
        "x-api-key": "<cacodes-personal-access-token>",
        "x-workspace-slug": "cacodes"
      }
    }
  }
}

Cursor (OAuth)

{
  "mcpServers": {
    "cacodes": {
      "url": "https://mcp.cacodes.com.br/http/mcp",
      "type": "http"
    }
  }
}

Environment variables

CaCodes (docs)	Env var	Legado (compat.)	Description
CaCodes API Key	CACODES_API_KEY	PLANE_API_KEY	Personal access token from CaCodes
CaCodes URL	CACODES_BASE_URL	PLANE_BASE_URL	https://tarefas.cacodes.com.br
Workspace	CACODES_WORKSPACE_SLUG	PLANE_WORKSPACE_SLUG	e.g. cacodes
AgentMemory	AGENTMEMORY_URL	—	Server-side memory (internal)

Memory-first

GSD + Ponytail are always-on after install. Call cacodes_get_project_context and memory_recall before scanning the codebase. Map context into AgentMemory via cacodes_trigger_reindex — not local .planning/ as source of truth.

Release

Published via .github/workflows/mcp-pypi-publish.yml on GitHub Release. Ponytail vendor snapshot is bundled under plane_mcp/bundled/ponytail.