pith
MCP server for task management that enables AI agents to read, create, update tasks, and track work sessions, allowing agents and humans to collaborate on the same task board.
README
<p align="center"> <h1 align="center">Pith</h1> <p align="center">Task management where AI agents and humans are equal teammates.</p> </p>
<p align="center"> <a href="#quick-start"><strong>Quick Start</strong></a> · <a href="#connect-your-ai-agent"><strong>Connect Your Agent</strong></a> · <a href="#cli"><strong>CLI</strong></a> · <a href="#web-ui"><strong>Web UI</strong></a> · <a href="#api"><strong>API</strong></a> · <a href="#ai-features"><strong>AI Features</strong></a> </p>
<p align="center"> <a href="https://opensource.org/licenses/MIT"><img src="https://img.shields.io/badge/License-MIT-blue.svg" alt="License: MIT"></a> <img src="https://img.shields.io/badge/node-%3E%3D20-brightgreen" alt="Node.js 20+"> <img src="https://img.shields.io/badge/PostgreSQL-16%2B-336791" alt="PostgreSQL 16+"> <img src="https://img.shields.io/badge/MCP-Native-purple" alt="MCP Native"> </p>
<p align="center"> <img src="docs/screenshots/pith-demo.gif" alt="Pith — Board view, list view, task details, and agent activity" width="720"> </p>
AI coding agents do real engineering work — they read specs, write code, fix bugs, and submit PRs. But they're blind to the project's task board. Every context switch requires a human to copy-paste task details, update statuses, and relay decisions.
Pith fixes this. It's an open-source task management system where agents can read tasks, update progress, log work, and create sub-tasks — the same way humans do, but through APIs and MCP instead of a browser.
Quick Start
Docker (recommended)
git clone https://github.com/SiluPanda/pith.git
cd pith/docker
docker compose up
The API is ready at http://localhost:3456. Check it with curl http://localhost:3456/health.
Without Docker
Requires Node.js 20+ and PostgreSQL 16+.
git clone https://github.com/SiluPanda/pith.git
cd pith
cp .env.example .env
# Edit .env — set DATABASE_URL and JWT_SECRET
npm install
npm run db:migrate
npm run dev
Create your first user and project
# Seed sample data (creates admin user, project, and example tasks)
npm run db:seed
# Or use the API directly:
curl -X POST http://localhost:3456/api/v1/users \
-H "Content-Type: application/json" \
-d '{"name": "Admin", "email": "admin@example.com", "role": "admin"}'
# Save the apiKey from the response — it's shown only once
Connect Your AI Agent
Pith is a native MCP tool server. Any MCP-compatible client — Claude Code, Claude Desktop, Cursor, Windsurf, or custom agents — works out of the box.
Setup
Add to your MCP client config (e.g. .mcp.json):
{
"mcpServers": {
"pith": {
"command": "npx",
"args": ["-y", "@pith/mcp-server"],
"env": {
"PITH_URL": "http://localhost:3456",
"PITH_API_KEY": "kb_your_api_key_here"
}
}
}
}
What your agent can do
| Tool | What it does |
|---|---|
list_tasks |
Query tasks with filters — status, priority, assignee, labels, free-text search |
get_task |
Get full task details including comments, sub-tasks, and activity history |
create_task |
Create a task with title, description, priority, labels |
update_task |
Change status, priority, assignee, or any other field |
add_comment |
Post a Markdown comment on a task |
create_subtasks |
Break a task into sub-tasks (up to 20 at once) |
search_tasks |
Full-text search across all tasks |
get_my_tasks |
See what's assigned to the current agent |
get_context |
Get rich context — parent task, siblings, recent activity |
start_session / end_session |
Track work sessions with summaries |
Your agent also gets read access to live resources:
pith://project/{slug}/board— Current board state by status columnpith://project/{slug}/backlog— Full backlog with prioritiespith://task/{id}/context— Complete task context for deep workpith://user/{id}/workload— Current assignment load
Example: autonomous coding agent workflow
1. Agent calls get_my_tasks → sees "Fix auth middleware" assigned to it
2. Agent calls get_context → reads task details, parent task, recent comments
3. Agent calls start_session → begins tracked work session
4. Agent writes code, runs tests → (happens outside Pith)
5. Agent calls update_task → moves status to "in_review"
6. Agent calls add_comment → posts summary + PR link
7. Agent calls end_session → records what it accomplished
No human had to copy-paste context or update the board.
CLI
Install globally or use via npx:
npx @pith/cli init --url http://localhost:3456 --key kb_your_key --project my-project
Managing tasks
# List tasks with filters
pith task list --status todo --priority P0
# Create a task
pith task create "Implement rate limiting" --priority P1 --labels security,api
# View task details with comments and activity
pith task show <task-id>
# Update status
pith task update <task-id> --status in_progress
# Add a comment
pith task comment <task-id> "Started work, ETA 2 hours"
# Search across all tasks
pith search "authentication bug"
Agent sessions
pith session start --name "Claude Code" --tasks <task-id-1>,<task-id-2>
# ... do work ...
pith session end <session-id> --summary "Fixed auth bug and added tests"
pith session list
Machine-readable output
Every command supports --json for piping into scripts:
pith task list --status todo --json | jq '.[].title'
Web UI
Pith includes a lightweight web interface at http://localhost:5173 (dev mode) or served by the API in production.
- Board view — Kanban-style columns by status
- List view — Sortable table with filters
- Task detail — Full context with comments, sub-tasks, and activity timeline
- Agent activity feed — Review what your AI agents have been working on
- Session review — Inspect individual agent work sessions
Start the dev server:
cd packages/web
npm run dev
API
Full REST API with OpenAPI/Swagger documentation at /docs (development mode).
Authentication
Every request requires a Bearer token — either an API key or a JWT access token:
# Using an API key
curl -H "Authorization: Bearer kb_your_api_key" http://localhost:3456/api/v1/projects
# Using JWT (get tokens via login)
curl -X POST http://localhost:3456/api/v1/auth/login \
-H "Content-Type: application/json" \
-d '{"email": "admin@example.com", "apiKey": "kb_your_api_key"}'
Key endpoints
| Method | Endpoint | Description |
|---|---|---|
GET |
/api/v1/projects |
List projects |
POST |
/api/v1/projects |
Create project (admin) |
GET |
/api/v1/projects/:slug/tasks |
List tasks with filters |
POST |
/api/v1/projects/:slug/tasks |
Create task |
GET |
/api/v1/tasks/:id |
Get task with full context |
PATCH |
/api/v1/tasks/:id |
Update task fields |
POST |
/api/v1/tasks/:id/comments |
Add comment |
POST |
/api/v1/tasks/:id/subtasks |
Batch create sub-tasks |
GET |
/api/v1/search?q=... |
Full-text search |
POST |
/api/v1/sessions |
Start agent session |
GET |
/api/v1/projects/:slug/analytics |
Project analytics |
Full reference: docs/api-reference.md
Roles
| Role | Can do |
|---|---|
| admin | Everything — manage users, projects, webhooks, tenants |
| member | Create/update tasks, add comments, view projects |
| agent | Same as member — designed for AI agent API keys |
Webhooks
Get notified when things happen in your project:
curl -X POST http://localhost:3456/api/v1/projects/my-project/webhooks \
-H "Authorization: Bearer kb_admin_key" \
-H "Content-Type: application/json" \
-d '{"url": "https://your-server.com/webhook", "events": ["task.created", "task.updated"]}'
Events: task.created, task.updated, task.deleted, comment.created, session.started, session.ended, project.created, *
Payloads are signed with HMAC-SHA256 via the X-Pith-Signature header.
Slack & Discord notifications
curl -X POST http://localhost:3456/api/v1/projects/my-project/notifications \
-H "Authorization: Bearer kb_admin_key" \
-H "Content-Type: application/json" \
-d '{"provider": "slack", "name": "dev-channel", "webhookUrl": "https://hooks.slack.com/...", "events": ["task.created", "task.status_changed"]}'
AI Features
AI features are optional. Pith works fully without any AI model configured. When configured, AI enhances — never blocks — your workflow.
Configure a provider
# Via CLI
pith config set ai.provider anthropic
pith config set ai.model claude-sonnet-4-20250514
pith config set ai.apiKey sk-ant-...
# Or via environment variables
export PITH_AI_PROVIDER=anthropic # anthropic, openai, google, groq, ollama
export PITH_AI_MODEL=claude-sonnet-4-20250514
export PITH_AI_API_KEY=sk-ant-...
export PITH_AI_BASE_URL= # optional, for self-hosted models
Supports any provider via Vercel AI SDK: Anthropic, OpenAI, Google, Groq, OpenRouter, and Ollama for local models.
What AI can do
- Decompose tasks — Break a large task into actionable sub-tasks with estimates
- Triage — Auto-suggest priority and labels for new tasks
- Context assembly — Build a briefing document with key points and risks for an agent starting work
- Effort estimation — Suggest time estimates based on historical project data
- Sprint summaries — Generate project summaries from activity data
- Duplicate detection — Flag similar tasks using pg_trgm similarity
# Decompose from CLI
pith task decompose <task-id>
# Get AI-assembled context
pith task context <task-id>
# Via API
curl -X POST http://localhost:3456/api/v1/ai/triage \
-H "Authorization: Bearer kb_..." \
-H "Content-Type: application/json" \
-d '{"title": "Fix memory leak in worker pool", "description": "Workers are not being cleaned up..."}'
Multi-Tenant Mode
Pith supports multi-tenant SaaS deployments with isolated workspaces:
curl -X POST http://localhost:3456/api/v1/tenants \
-H "Authorization: Bearer kb_admin_key" \
-H "Content-Type: application/json" \
-d '{"slug": "acme-corp", "name": "Acme Corp", "plan": "pro"}'
Each tenant gets configurable user and project limits.
Configuration Reference
Environment variables
| Variable | Required | Default | Description |
|---|---|---|---|
DATABASE_URL |
Yes (production) | postgres://postgres:postgres@localhost:5432/pith |
PostgreSQL connection string |
JWT_SECRET |
Yes (production) | dev fallback | Secret for signing JWT tokens |
PORT |
No | 3456 |
API server port |
HOST |
No | 0.0.0.0 |
API server bind address |
CORS_ORIGIN |
No | http://localhost:5173 |
Allowed CORS origins (comma-separated) |
LOG_LEVEL |
No | info |
Log level: debug, info, warn, error |
PITH_AI_PROVIDER |
No | — | AI provider name |
PITH_AI_MODEL |
No | — | AI model identifier |
PITH_AI_API_KEY |
No | — | AI provider API key |
GITHUB_WEBHOOK_SECRET |
No | — | Secret for verifying GitHub webhook signatures |
Project Structure
packages/
core/ Shared types, Zod schemas, constants
db/ PostgreSQL schema, migrations, seed data (Drizzle ORM)
server/ REST API — Fastify, auth, RBAC, webhooks, analytics
ai/ AI integration layer (Vercel AI SDK, provider-agnostic)
mcp-server/ MCP tool server (stdio + HTTP transports)
cli/ Command-line interface (Commander.js)
web/ Web UI (React + Vite)
Contributing
See CONTRIBUTING.md for development setup and guidelines.
git clone https://github.com/SiluPanda/pith.git
cd pith
npm install
npm run db:migrate
npm test # 146 tests
npm run dev # Start API server with hot reload
License
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.