Procni

Procni is a local-first, open-source developer tool that watches what you've actually built (via git) and turns that work into publishable content — social posts, technical write-ups, build-in-public updates, screenshots, and eventually demos and video. It runs entirely on your machine with your own AI provider key (BYOK) — no Procni account, no cloud dependency for the core loop.

procni init
procni dev

opens a local dashboard where recent development work becomes reviewable Content Opportunities you can turn into drafts, edit, and copy out.

The full architecture and phased roadmap live in [.claude/plans — not committed]; the authoritative day-to-day source of truth for what's built and what's next is CHECKLIST.md.

Build-it-yourself: the module README chain

This project is documented so that a developer new to building this kind of tool (local-first CLI + server + AI pipeline) can read every module's README in order and understand not just what it does but why it's built that way — including the alternatives that were rejected. Read them in this order:

1. packages/shared — shared DTOs and error types used everywhere else.
2. packages/core/src/config — loading config.toml, global vs. project settings.
3. packages/core/src/db — SQLite schema, migrations, the repository pattern.
4. apps/cli — the procni command line: init, dev, doctor.
5. apps/server — the local REST/SSE server the web UI talks to.
6. web — the React SPA that renders opportunities and drafts.
7. packages/core/src/git — shelling out to real git safely.
8. packages/core/src/context — turning git history into a redacted, size-bounded DevelopmentContext.
9. packages/ai — the AI provider abstraction, structured output, key storage.
10. packages/core/src/pipeline — the staged Context→Opportunity→Draft pipeline.
11. packages/core/src/jobs — the SQLite-backed job queue and worker.

Each README is added in the same commit as the code it documents — as new modules land, this list grows.

Status

Phases 0, 1, 2, and 3 (Review experience) are complete: procni analyze finds Content Opportunities in real git history, and the dashboard's platform tabs (X/Twitter, LinkedIn, Blog, Changelog) each generate and hold their own independent draft — a developer can edit any of them, approve, dismiss/restore an opportunity, browse and restore past revisions per platform, and Copy/Export .md, with a live length counter (against a per-platform target a developer can configure themselves) and generation progress throughout. A right-hand slide-over shows the real source diff a draft is (or would be) grounded in, and lets a developer select specific files or highlight a code block to spin off into a brand new opportunity of their own choosing — not just whatever the automated analysis surfaced. A project-level voice/tone style guide shapes every future draft. The whole UI uses real icons (Hugeicons), not emoji. BYOK spans Gemini, Anthropic, OpenAI, OpenRouter, and local Ollama, resolved automatically from whichever key (or PROCNI_OLLAMA_BASE_URL) is actually configured. All of it verified against Procni's own repository with a real Gemini key, fulfilling the "Procni documents Procni" milestone. See CHECKLIST.md for exactly what's built and where to find it; Phase 4 (Screenshots) is next.