Open-source Irrigator adds crash recovery and observability to coding agent swarms
volatilityfund · reddit · 2026-11-17
The author released Irrigator, an MIT-licensed Go project for running coding agent swarms locally. Beyond parallelism, it tackles tracking agent work, recovering from failures, and knowing what actually completed.
- Persistent workflows, isolated Git worktrees, and explicit recovery states
- Interrupted agents aren't blindly replayed: state is preserved for inspection first
- Inspect runs, retry failed tasks, resume interrupted work
- Supports Claude Code, Codex, Antigravity, and Herdr as runtimes
- Runs locally with a mock demo requiring no API keys
GitHub: grainulation/irrigator. The author invites feedback on handling durability in multi-agent workflows.
More from coding & agent
- Developer finds his agent autonomously built a playground to explore its own generated content — evilrabbit_ · 2026-11-17
- Voice agent passes 15 metrics but customers still complain — what's not being measured? — georgeamine · 2026-11-17
- Compilers are piles of human heuristics tuned for the average case — yours isn't average — hargup13 · 2026-11-17
- Google's modern-web-guidance lifts coding agents' best-practice pass rate from 52% to 92% — victor_explore · 2026-11-17
- Orchestrating local LLMs: assign skills per model, let a third one debate and verify — gnukeith · 2026-11-17
- 435K LOC deep dive: a hands-on manual for LangChain 1.5, LangGraph 1.2 and Deep Agents — ghumare64 · 2026-11-17