# Pith

> Pith is where vibe-coded projects go to grow up.

Push your code, get it live, and let Pith help your agent spot problems before they bite you. Hosting, diagnostics, task tracking, and project memory - your agent learns the tools as it uses them.

## What you get

- **Hosting** - Deploy from a git repo with one tool call. Buildpacks, containers, TLS, DNS - handled.
- **Diagnostics** - Your agent asks questions about app health in plain language and gets answers. Log ingestion and pattern detection included.
- **Tasks** - Track work with dependencies, relationships, and attachments. Your agent manages its own backlog.
- **Memory** - Semantic search over everything your agent has learned. Pin the important stuff, search the rest.
- **Git** - Self-hosted repos with SSH access. Push, diff, log - the basics.
- **Organizations** - Multi-tenant from day one. Teams, projects, permissions.

## Quick install

> Install the Pith MCP server. Read https://usepith.dev/llms.txt for full instructions, then follow the "Getting started" steps to register and connect.

Paste the prompt above into any AI agent. It reads `llms.txt` and handles the rest.

## Getting started

*Humans: just [sign in](/login/). The steps below are for connecting an AI agent.*

1. Ask the human for their email address.
2. POST to https://usepith.dev/register/ with JSON body:
   {"email": "<their email>", "agent_name": "<your name>"}
   Response: {"status": "pending", "request_id": "<uuid>"}
3. Tell the human to check their email and click the approval link.
4. Poll GET https://usepith.dev/register/status/<request_id>/ every 5 seconds
   until the response status is "approved". The approved response includes:
   {"status": "approved", "token": "...", "mcp_config": {...}}
5. Add the returned mcp_config to your MCP configuration file
   (e.g. .mcp.json for Claude Code, or the equivalent for your client).
6. Ask the human to restart the session so the new MCP server loads.

## Connecting

Pith talks MCP over streamable HTTP. One endpoint, no REST API.

- **Endpoint**: /mcp/ (streamable HTTP, stateless)
- **Auth**: agent token (Authorization: Token <token>) or OAuth 2.1 (RFC 9728)
- **Discovery**: GET /.well-known/mcp.json

## Docs

- [Full tools reference](https://usepith.dev/llms-full.txt): All MCP tools with parameters and examples
- [Install guide](https://usepith.dev/install.md): Setup for humans and agents
- [MCP discovery](https://usepith.dev/.well-known/mcp.json): Machine-readable server metadata

## Optional

- [Install config (JSON)](https://usepith.dev/install.json): Machine-readable install configuration