Durable runs are now GA

Agents that don’t stall.

Warrick is the orchestration layer for AI agents. Durable runs, automatic retries, step-level replay, and a trace for every decision your agent made — so you stop rebuilding the same scaffolding on every project.

$ npm i @warrick/sdk
await run(researchAgent, {
  retries: 5,
  checkpoint: "each-step",
  onStall: "resume"
})
→ run_8fk2 · 41 steps · 2 retries · resumed once ✓
Orchestrating agents atCinderHalyardParityThresherRoost

Why teams switch

Your agent framework stops at the happy path.

Durable by default

Every step is checkpointed. A crashed worker, a rate limit, a 3am deploy — the run picks up where it stopped instead of starting over.

Replay any decision

Step-level traces with the exact prompt, tool call, and response. Scrub back to step 14 and re-run it against a different model.

Stall detection

Agents don't crash, they loop. Warrick notices when yours is going in circles and resumes, escalates, or halts on your rules.

Pricing

Free while you’re figuring it out.

Usage-based after that. No seat pricing, no sales call.

Free

$0
1 concurrent run
500 steps / month
7-day trace retention
Community Discord
Most popular

Pro

$20/ month
Unlimited concurrency
50k steps / month
90-day traces + replay
Shared Slack channel

first 50 pro coin included

No. 36 of 50 · 15 left

Team

$99/ month
Everything in Pro
Self-hosted workers
SSO and SAML
99.9% uptime SLA

first 25 team coin included

No. 4 of 25 · 22 left

2 developers subscribe to Warrick