Raenil

The idea

Most trackers are built for humans coordinating with humans. Raenil is built for a different shape of work: one person directing AI agents, then reviewing what they produced. Its job isn’t to replicate Linear or Jira — it’s to keep an honest record of what was done, how it was done, by whom (human or AI), and the artifacts left behind.

Architecture

A single Go binary serves a SvelteKit SPA and talks to PostgreSQL — embedded SQL migrations apply on boot, so deployment is one container. The distinctive piece is a built-in MCP server: AI coding agents connect over a bearer token and drive the board directly — create issues, set acceptance criteria, link the exact commits that implemented a change, write engineering artifacts, and move work through states.

Every actor action — human via session, AI via token — is attributed and persisted to an activity log. That log is the backbone the rest of the product is built on.

Engineering highlights

• Actor-attributed activity log as the source of truth; history survives issue deletion via snapshotted keys/titles.
• Code-links, done-when checklists, and coverage surfacing so a finished issue points at its commits, its acceptance criteria, and its artifacts.
• Inbox review queue — the human lands on exactly what the AI moved to In Review while they were away, with approve / bounce actions that map to the workflow gate.
• MCP tool surface teaching agents the “finishing an issue” ritual: move state → link commit → write artifact → check criteria.

Runs behind a Cloudflare Tunnel with edge auth; four-level hierarchy (Project → Epic → Issue → Sub-issue); Web Push for review notifications.