Quick Start
From install to your first plan in about five minutes. You bring the AI coding tool. PAPI brings the memory and the map.
Sign in
Go to getpapi.ai/login and sign in with GitHub or email. This is the account your connector authenticates against, so use the same one in both places.
Connect PAPI to your AI tool
PAPI runs as a remote MCP connector at mcp.getpapi.ai. Nothing to install or host. In Claude Code, from your project root:
The server is added but not yet authenticated. Run /mcp, choose papi, and pick Authenticate. A browser tab opens; sign in with your getpapi.ai account and you are connected.
mcp.json. Copy your ready-made snippet from getpapi.ai/settings so it carries your real values.Set up your project
Ask your AI tool to run setup. PAPI reads your repo, writes a Product Brief, seeds a few Active Decisions, and drops a CLAUDE.md with your conventions. One call, and your project has a memory.
setup scans your existing structure and builds context-aware artifacts instead of starting from a blank page.Plan your first cycle
Now ask PAPI to plan. It sorts your backlog into a focused cycle and writes a BUILD HANDOFF for each task: scope, acceptance criteria, files to touch, and an effort estimate. The handoff is what lets the next step run without you re-explaining anything.
Build the first task
Pick a task and build it. PAPI hands your AI the full spec, creates a branch, and captures a build report when the work lands.
That is the loop. From here it is orient, plan, build, review, release, cycle after cycle, each one a little sharper than the last.
orient. It is the “where am I and what is next” call, and it opens every session.