Skip to main content
ALPHA 1

Run your build and test workflows locally.

Define your commands in YAML and run them on the host or in Docker containers. Loom records job and step results in linked files that you or your agent can inspect when a command fails.

Alpha 1 is an evaluation-stage release. Read the support boundary before relying on a capability or compatibility claim.

Run Hello LoomSee how Loom works
1. Receiptlogs_dir2. Pipeline manifestfailing_job_manifest_path3. Job manifestStep or system pointer4. events.jsonlOutput to investigateFor runs with runtime artifacts

Alpha 1

evaluation-stage maturity

Local

current execution scope

Linux

schema v1 target

From a local workflow to its output

Define a host job

Workflow YAML
version: v1
stages: [ci]

test:
  stage: ci
  target: linux
  script:
    - go test ./...

For a Go project with Go installed, save this as .loom/workflow.yml. Run loom check, then loom run --local. Replace the script with your project's test command.

Inspect a failed run

Runtime files
receipt (logs_dir)
  → pipeline/manifest.json
  → jobs/<job_id>/manifest.json
  → jobs/<job_id>/user/execution/script/<NN>/events.jsonl

For a script failure, follow the receipt and manifests to the failed entry's output. If setup fails before runtime files exist, start with the terminal error.

Why use Loom

Job and step diagnostics

Use manifest pointers to locate output from the failed script entry or system section.

Learn more →

Structured runtime artifacts

Read receipts, manifests, and event streams using a documented layout. IDs, timings, paths, and output vary between runs.

Learn more →

Local host and Docker execution

Run job scripts with host tools or a Docker image, with workflow settings for caching, secrets, and templates.

Learn more →

How it works

1

Author

Write a workflow in .loom/workflow.yml.

2

Validate

Run loom check to validate its structure.

3

Execute

Run loom run --local.

4

Diagnose

Inspect the output linked from the job manifest.

Inspect results with an agent

Use the same run files from an editor, a script, or an agent. File sizes depend on the workflow and its output.

Select the output to inspect

An agent can read the receipt and manifests to locate a failed script entry or system section.

Use the recorded paths

Follow the paths in each run's manifests. The selected output still needs interpretation.

Read JSON and JSONL

Receipts and manifests contain metadata and results; event streams contain execution events and command output.

Start here

Choose a guide for the task you want to try.

Run Hello Loom

Run a first workflow, inspect the artifacts it creates, and diagnose one intentional failure end to end.

See how Loom works

Learn how Loom validates and executes workflows, and where it records the results.

Diagnose a failing run

Find the relevant job or step output, including where to start when setup fails.

Try the Hello Loom example

Run a small workflow, introduce a failure, and inspect the files Loom writes.

Run Hello Loom