Can we just
keep it super simple?
Free for individuals, academics, and teams. We are no longer selling the per-seat plans — roar and hosted GLaaS stay free to use, and if you need something beyond that, the answer is a conversation rather than a price list.
- roar CLI
- Hosted GLaaS · private lineage
- 1 writer seat · 1-year retention
- 5 projects · BYO node
- SSO, SAML, audit logs
- Role-based access control
- Fully customizable policies
- Dedicated support & SLA
- White-glove onboarding
- Procurement & security review
Compare tiers
The honest version. If something isn't on this table, ask us.
| Free | Enterprise | |
|---|---|---|
| Capture & lineage | ||
| roar CLI | ✓ | ✓ |
| GLaaS registry | hosted | hosted |
| Lineage retention | 1 year | unlimited |
| Private lineage | ✓ | ✓ |
| Live Lineage | — | ✓ |
| Integrations | ||
| Experiment trackers (W&B) | ✓ | ✓ |
| Cloud capture (AWS, S3) | ✓ | ✓ |
| Multi-node capture (Ray) | — | ✓ |
| Slack | — | ✓ |
| Coordination | ||
| Reader seats | — | unlimited · free |
| Projects | 5 | unlimited |
| treqs CLI | ✓ | ✓ |
| Training requests | ✓ | ✓ |
| Customizable policies | — | full |
| Access control (RBAC) | — | ✓ |
| Orchestration | ||
| Compute queue | ✓ | ✓ |
| BYO nodes (dedicated) | 1 | unlimited |
| Lambda on-demand | — | ✓ |
| Platform | ||
| API & webhooks | — | ✓ |
| Cost & compute attribution | — | ✓ |
| SSO / SAML, audit logs | — | ✓ |
| Support | ||
| Support | community | dedicated + SLA |
| Onboarding | self-serve | white-glove |
Free for academic use.
Researcher, grad student, or lab trying to make your work reproducible — TReqs is free, with no time limit.
academic@treqs.ai →Questions you might still have
The ones we get asked most. If yours isn't here, email us.
Do I need a credit card for the free tier?
No. roar and local GLaaS run entirely on your machine. No signup required, no card needed.
What counts as a "writer seat"?
Anyone who runs roar run, files training requests, approves them, or pushes lineage to the team registry. Viewers who only need to read results, browse lineage, or track costs are free reader seats — they can't run or approve.
Can agents file training requests?
Yes. Agents use the treqs CLI — similar to how gh works for GitHub — under your credentials. No separate seat: the agent acts as you, consuming your writer seat. Only humans can approve requests (unless you configure a policy otherwise).
Do you have a nonprofit or startup discount?
Academic and research use is free — see the note above. Pre-seed startups: talk to us, we're flexible.
Does GLaaS store my models or training data?
No. GLaaS stores lineage — what ran, what inputs were read, what outputs were written, and the environment they ran in. Your models, datasets, and checkpoints stay where they are. GLaaS is a registry, not a data store.
Evaluating this for an ML platform team?
There's a dedicated overview: where roar and GLaaS sit in your stack, what leaves your machines, and the build-vs-buy case. For platform teams →
Produce a lineage
in 30 seconds.
Free to use. If you need more than that, come talk to us.