Question 1

What is PAPI?

Accepted Answer

PAPI (Persistent Adaptive Project Intelligence) is a planning and memory layer for AI-assisted software development. It gives AI coding tools like Claude Code, Cursor, and Windsurf persistent project memory through structured cycles of plan, build, and review, so every session builds on the last instead of starting from scratch.

Question 2

How is PAPI different from CLAUDE.md or project context files?

Accepted Answer

CLAUDE.md and similar files are static. PAPI maintains live, structured project state across sessions: tasks, Active Decisions with confidence levels, build reports, review verdicts, strategy reviews every five cycles, and learnings that feed the next plan. The methodology is opinionated and gets smarter over time.

Question 3

Which AI coding tools does PAPI work with?

Accepted Answer

PAPI works with any MCP-compatible AI tool, including Lovable, Bolt, Replit, Cursor, Claude Code, Windsurf, and ChatGPT. Setup is pasting one connector link (or one command in CLI tools) and takes about a minute. The same project memory is shared across every tool, so you can switch tools or subscriptions without losing context.

Question 4

How much does PAPI cost?

Accepted Answer

PAPI is currently free in alpha. The MCP server and hosted dashboard at getpapi.ai are free to use. Future paid tiers will likely focus on advanced intelligence features (deeper strategy reviews, cross-project learning, hosted analytics) while the core management loop (cycles, builds, reviews) stays free.

Question 5

What is the cycle methodology?

Accepted Answer

A PAPI cycle is plan, build, review, release. The plan tool proposes prioritised tasks. build_execute creates a feature branch and a structured handoff your AI can implement against. review_submit records verdicts. release merges and rolls forward. Every five cycles, strategy_review analyses what is working and recommends direction changes.

Question 6

What makes PAPI different from my .md documents?

Accepted Answer

An .md doc is prose the AI has to keep honest by discipline. PAPI is state that updates itself as a side effect of working. It is structured (tasks, statuses, cycles, decisions and build reports, all queryable) instead of unstructured prose, uses scoped retrieval so one orient call returns just the current state instead of loading a file that grows without limit, enforces the plan, build, review, release loop with gates, keeps a decision registry with supersede history, and compounds by feeding effort estimates, scope accuracy and learnings into the next plan. The same state is available from Claude Code, Cursor, or any MCP client.

Question 7

Does PAPI add token usage, or does it reduce it?

Accepted Answer

PAPI adds zero inference cost: it makes no LLM calls of its own, needs no API key or credits, and the reasoning happens inside the subscription you already pay for. Tool responses do take up context, but they replace a heavier pattern. Markdown planning docs grow without limit and load in full every session, while PAPI responses stay about the same size no matter how old the project gets. Session start is one scoped orient call, handoffs are assembled once at plan time, and decision memory avoids re-arguing settled decisions. The result is a small fixed cost per call against an unbounded cost that grows every session, and the crossover favours PAPI within weeks on any live project.

Your AI can write the code. PAPI keeps the project together.

AI doesn’t lose your code. It loses the plot.

PAPI isn't an app you open. It works where you work.

Plan that compounds

Review before it lands

Use the AI that's best for the job.

Set up once. Then the same four steps, every time.

Every few cycles, PAPI steps back: are you still building the right thing?

Your project remembers why.

Run the workflow you choose.

Planned, built and released on PAPI.

Flat price. No metering. No credits.

Point your AI at PAPI.