Glossary
Every word PAPI uses on this site, defined once in plain language.
MCPthe plug between PAPI and your AI tool#
MCP stands for Model Context Protocol, a standard that lets AI tools like Claude Code, Cursor, or Codex connect to outside services and use them mid-conversation. PAPI is an MCP server: you add its address to your AI tool once, and from then on your AI can read and update your project while you chat. Nothing to install, no separate app to run.
Cycleone full lap: plan, build, review, release#
A cycle is the unit PAPI works in. You plan a small batch of tasks, build them, review what came back, and release the finished work as a version. Then the next cycle starts, sized from what the last one actually delivered.
BUILD HANDOFFthe work order your AI builds from#
A BUILD HANDOFF is a self-contained brief for one task: the goal, what is out of scope, how to check the work is done, the likely risks, and the files probably involved. When a task starts, PAPI hands your AI the brief, so it builds from instructions rather than guesswork. A good handoff is complete enough that an AI with no history on the project can run it.
Cold handoffa handoff run with no prior knowledge of the project#
A handoff runs cold when the AI executing it has never seen your project: a fresh chat, a different tool, a teammate’s session. Because a BUILD HANDOFF carries everything the task needs, the work still comes out right. This is the bar PAPI holds every handoff to.
Effort points (pts)task size, in comparable units#
Every task gets a size estimate, from XS to XL, which maps to points. Points put tasks and cycles on one scale, so a two point fix and an eight point feature can sit in the same plan and still add up honestly. The number matters less than how it compares to what the task actually took.
Velocityhow much a project really ships per cycle#
Velocity is the number of effort points a project delivers in a cycle, averaged over recent cycles. PAPI uses it to size the next plan, since what recently shipped is the best predictor of what fits next. It is a reference, not a target: chasing the number only inflates estimates.
Scope accuracyhow close estimates land to reality#
After every build, the actual effort is recorded beside the estimate. Scope accuracy is how often the two match, and in which direction they miss. If a project keeps over-scoping by a few points, the next plan is sized with that bias in mind.
Active Decisiona standing call your AI checks before acting#
An Active Decision records a choice that should keep binding future work: an architecture pick, a naming rule, a direction you settled. It carries the reasoning behind the call and a confidence level that rises or falls with evidence. Your AI checks these before making a related choice, which is how a decision made in one tool in March still holds in another tool in July.
Decision replaystep back through how a call was made#
Every recorded decision keeps its reasoning, its confidence, and each time it changed. Decision replay walks that history: what was decided, why, what evidence later shifted it, and what took its place. Superseded calls are never erased, so the trail stays complete.
Strategy reviewa scheduled zoom-out every few cycles#
Every five cycles, PAPI reads the accumulated build reports, decisions, and metrics and asks the bigger question: is the project still heading the right way? The answer comes back as recommendations that feed straight into the next plan. The zoom-out happens on schedule, not when someone remembers.
Driftthe slow slide away from what you meant to build#
Drift is what happens when small unchecked changes stack up: scope grows mid-task, a settled question gets re-decided in a fresh chat, and the plan stops describing the project you are actually building. PAPI counters it with a scope boundary in every handoff, decisions your AI checks before acting, and a strategy review on a schedule.
These are the short versions. The concepts guide goes deeper on decisions, handoffs, and cycles.