convez.ai
MCP Server
ItsJennyFiggy

plane-mcp

ItsJennyFiggy/plane-mcp

Ver no GitHub
InicialPluginScore · de 0 a 100

Token-efficient, agent-scoped Go MCP server for self-hosted Plane

Instalação

$npx convez-ai ItsJennyFiggy/plane-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
50%
Último commit
há 2 meses
Versão
v1.18.0
Issues abertas
0
Contribuidores
4

README

A tradução para o português deste README ainda está na fila.

plane-mcp

A custom Go-native Model Context Protocol (MCP) server that integrates directly with the Plane.so REST API. It exposes clean, token-efficient, name-resolved tools to AI agents for automated task tracking, progress reporting, and workspace planning.


🎯 Features & Scope

  • Stdio Transport: Communicates natively over standard I/O (stdio) to plug into Claude Desktop or container runtimes.
  • Tool Scoping Profiles: Restricts the exposed tool surface dynamically via PLANE_MCP_PROFILE=worker|planner|full (e.g., preventing worker agents from calling destructive or planning tools).
  • Token-Efficient Payload Serialization: Serializes work items into compact formats, resolves UUID-valued fields (states, assignees, labels) to human-readable names, converts description HTML to Markdown, and strips nulls to optimize LLM context usage.
  • Tier-1 Semantic Tools: High-level, developer-oriented operations:
    • find_my_work: Lists items assigned to the current caller.
    • get_work_item: Fetches detail/summary view of a ticket.
    • list_work_items: Lists work items in a project with optional filters.
    • search_work_items: Searches work items across the workspace by text query.
    • report_progress: Appends comments and transitions states safely in a single action.
    • submit_for_review: Attaches PR links and flags tickets for review.
    • create_task: Resolves labels/assignees by name and registers new tasks.
  • Tier-2 CRUD Tools (Gated): Low-level operations (update_work_item, list_comments, get_last_comment, comment lists, link creation) registered only under the planner or full profiles.
  • Off-Network Tunnel Ingress: Supports Cloudflare Access service tokens via headers to route requests securely to local/private Plane deployments.

🛠️ Tech Stack & Architecture

  • Runtime/Language: Go 1.26
  • Transport: Model Context Protocol Go SDK (stdio)
  • Build Packaging: Multi-stage distroless scratch builds (gcr.io/distroless/static-debian12)
  • CI/CD: Automatic linting/testing via GitHub Actions, and container publication to GHCR via OIDC.

🚀 Getting Started

Prerequisites

  • Go 1.26+ installed on host
  • A Plane.so workspace with a Personal Access Token (PLANE_API_KEY)

Local Setup

  1. Navigate to the repository directory:
    cd plane-mcp
    
  2. Configure local environment variables:
    # Create a local .env configuration file
    export PLANE_API_KEY="your-plane-api-key"
    export PLANE_BASE_URL="http://192.168.86.32:3355" # or https://plane.figgy.cloud
    export PLANE_WORKSPACE_SLUG="itsjennyfiggy"
    export PLANE_MCP_PROFILE="full"
    
  3. Install dependencies:
    go mod download
    
  4. Run the application locally:
    go run ./cmd/app
    

Running Tests

To run the local unit test suite and verify code coverage:

go test -v -race -cover ./...

🤖 AI Agent Guidelines

If you are an AI coding agent working in this repository:

  1. Always audit the local .agents/rules/ directory before making changes.
  2. Adhere strictly to .agents/rules/git_safety.md to prevent secrets exposure.
  3. Prior to presenting work or opening a PR, ensure all test suites pass and coverage gates are met (refer to .agents/rules/testing_standards.md).
  4. Follow the standard branching and PR lifecycle detailed in .agents/workflows/git-workflow.md.
Ver o original no GitHub

Relacionados