Skip to content

Potpie

Turn your codebase into a living context graph for AI agents.

Get started How it works

  • Quickstart

    Install the CLI and set up your first project in under 3 minutes.

    Get started

  • Context Graph

    Understand how Potpie builds a living project-memory layer for agents.

    Context Graph

  • Knowledge Graph

    Code-structural layer: AST-parsed nodes, semantic embeddings, and graph queries.

    Knowledge Graph

  • CLI Reference

    Full command reference — setup, integrations, graph workbench, skills, and more.

    Commands

  • Prebuilt Agents

    Ask, Build, Debug, and Spec — four agents powered by your graph.

    Prebuilt Agents

  • Custom Agents

    Define role, goal, and tasks — run them against your codebase.

    Custom Agents

Why Potpie?

Most AI coding tools operate on flat file context — they read what you paste or what fits in a window. Potpie is different:

Flat context Potpie context graph
Scope Current file + paste buffer Entire codebase + decisions + history
Structure None Functions, classes, dependencies as graph nodes
Semantics Raw text LLM-generated descriptions, tags, embeddings
Memory Session only Durable across agent sessions
Agent support Manual prompt stuffing Purpose-built query and traversal API

Install in one command

uv tool install potpie

python3 -m pip install --user potpie

Then run:

potpie setup --repo . --agent claude

Integrations

  • GitHub

    Repos, PRs, issues, reviews, source history

  • Linear

    Teams, issues, projects, documents

  • Jira

    Projects, issues, status, changelog

  • Confluence

    Spaces, pages, runbooks, decisions

Coding harnesses: Claude Code · OpenAI Codex · Cursor · OpenCode