Skip to main content

Alpha 1

Stop reconstructing failures.
Start following them.

Loom is the open-source workflow engine that turns every run into a deterministic evidence trail — so you land on the failing unit in three pointer hops, not three hours of log archaeology.

The run failed. Now what?

You open the logs. Thousands of interleaved lines. You ctrl-F for "error," find twelve matches across six jobs, then spend the next hour cross-referencing timestamps to figure out which failure actually mattered. By the time you hand the context to a teammate — or an agent — the trail is cold and nobody trusts the reconstruction.

The Loom difference

🎯
3 pointer hops to root cause

Receipt → manifest → events. Every run writes the same deterministic chain, so you never guess where to look.

📦
~2 KB to the failing unit

Structured JSON artifacts, not megabytes of interleaved text. Hand the same path to any session or contributor and they see the same evidence.

🤖
Agent-native artifacts

Deterministic, structured evidence that AI agents can parse and act on — no brittle log scraping, no prompt-stuffing raw output.

How it feels

✍️Author
Validate
🚀Run
🔍Diagnose

Write a workflow. Validate it before it burns compute. Run it locally with host or Docker providers. When something breaks, follow the pointer chain to the exact failing unit — no guessing, no reconstruction.

Alpha 1 — what's real today

Alpha 1 ships local execution with host and Docker providers, built-in caching, secrets, template composition, and the full pointer-first diagnostic chain. The artifact model is stable. Runtime fields may evolve between releases. Remote execution is planned — if you need it today, Loom isn't there yet. If you want a local workflow engine with a stronger failure-triage model, this is the right evaluation point.

Explore

Ready to follow the failure trail?
Hello Loom gets you to your first workflow, first artifacts, and first diagnosis in under five minutes.
Start with Hello Loom