Every menu and dialog across the four products, captured live from the
running portals. Each product is useful on its own and hands off to the next along the
PARR loop — Prepare, Act, Reflect, Review.
One task, end to end PARR in flight
A single task — "Add a /version endpoint" — driven live through the whole
pipeline: planned and decomposed, coded subtask by subtask, verified, and watched from the cockpit.
1 · Prepare — PFactory decomposes the plan into criteria and child issues; gates pass.2 · Act — AIFactory breaks it into subtasks: endpoint, version string, unit test.2 · Act — the live agent log as the coder works the repo.3 · Reflect — TFactory's verdict and report from a real run (PASS, with evidence).4 · Review — CFactory threads it as one work item with a live execution DAG.
PFactory Prepare
Planning Portal — the plan board (Plans ready → Human Review → Done)New plan — ingest a document or paste textFiles — the project workspaceIndex & Memory — AI-discovered project knowledgeMCP — connected tools and serversSkills — reusable planning skills
AIFactory Act
Tasks — the PARR board (Plan → Code → Review → Done)New task — the build wizardSettings — agent profiles and per-phase model selectionChat — converse with the coding agentTerminal — the live agent shellFiles — the worktree browserWorktrees — isolated build workspacesIndex & Memory — codebase knowledgeChangelogGitHub IssuesGitHub PRsMCP — connected toolsSkills
TFactory Reflect
TFactory Pipeline (Plan → Generate → Execute → Report)New test taskNew test planCloud Reports — live deployment test runsVisual Reports — browser screenshots and recordingsTest PlansFilesChangelogGitHub PRsMCP — connected toolsSkills
CFactory Review
Mission Control — the cockpit over the whole pipelineCopilot — the advise-and-confirm assistantPipeline — plan → code → test across every siblingActive tasks — in-flight work, every factoryAudit — HMAC-chained activity and confirmed actionsTokens & Cost — LLM usage across the pipelineServices — upstream health and endpointsSettings