How PAPI works

PAPI gives your AI assistant a structured loop to run with you, and the memory to make each pass through the loop smarter than the last.

The cycle

plan → build → review → release → (next plan)

• plan looks at your backlog, your past cycles, and your active decisions, then proposes a cycle: a small set of right-sized tasks, each with a build handoff.
• build is where your assistant implements a task. Starting a build creates a branch and returns the handoff; completing it records a build report: actual effort, surprises, and anything discovered along the way.
• review records your verdict on each build: accept, request changes, or reject.
• release closes the cycle, merges the work, and tags it. Then the next plan starts from everything the last cycle learned.

Build handoffs

A handoff is a self-contained spec for one task: why now, what to do, what NOT to do, acceptance criteria, likely files, and a pre-mortem of what could go wrong. Handoffs are written so an assistant can execute them without extra context. They persist with the task, so a different tool (or a teammate's assistant) can pick the task up cold.

Decisions

Architectural and strategic choices are recorded as Active Decisions with a confidence level. Your assistant checks them before making related choices, and supersedes rather than overwrites them when direction changes, so the history of why stays intact.

Strategy reviews

Every few cycles, PAPI steps back: are the cycles pointing the right direction? What recurring friction should become a task? The output feeds the next plan, so course corrections happen on evidence, not vibes.

Where things live

• Your AI tool is the command surface. Plans, builds, and reviews all happen in conversation, through MCP tools.
• The dashboard at getpapi.ai is the review and visibility surface: board, cycles, decisions, briefs, and your team.
• Your data is stored per-project and per-user in PAPI's hosted backend. The remote transport authenticates every request; project data is scoped to your account. Team members you add to a project can see its board, decisions, and brief read-only.

What PAPI is not

• Not a code generator. Your AI tool writes the code; PAPI keeps the plan, the memory, and the loop.
• Not a ticket system you maintain by hand. The board is written and read by the tools as a side effect of working.
• Not autonomous. The loop pauses for your decisions at plan approval and review. You stay in control; that is the point.