AI analysis grounded in the code graph — computed facts, not vibes · 2026-10-05T03:03:24Z
OpenRig is a TypeScript harness that orchestrates Claude Code and Codex coding agents as a single managed system, configured via YAML "rig" definitions and operated through a CLI (rig setup, rig commands) and a terminal UI. It targets developers who want to run multiple AI coding agents against a repository as a persistent, addressable team rather than as ad-hoc terminal sessions, with a daemon (createDaemon in packages/daemon/src/startup.ts) coordinating agent lifecycle, a TUI for inspecting seats/runtime/context state, and an optional React-based UI (components like RigGraph, TopologyTableView, Feed) for visualising the rig as a graph or table.
The 624-star jump aligns with a dense release cadence — five releases in four days (v0.5.17 through v0.6.3, 2026-09-27 to 2026-09-30) — and release titles signalling a deliberate push toward reliability for new users ("works for strangers", "first-use repairs", "first-project and recovery fixes"). The commit history backs this up: the last 30 days are dominated by targeted fix(daemon), fix(cli), and fix(tmux) commits addressing concrete failure modes (UTF-8 chunk decoding, Windows ps liveness, SSE subscription cleanup), suggesting the project crossed a usability threshold around v0.6.0 that likely drove adoption. The README's "real recording" demo and installer polish (Bun support in v0.5.17) also point to a conscious onboarding push, though the graph facts alone can't confirm external promotion (e.g. the linked X account) as a growth driver.
What changed recently, how it's actually built (from the code graph), and whether you should care. Free account — no card, no spam.