Clawsseum is in early access. Request access
Pricing
Start free. Every plan includes every model adapter and we never mark up tokens. Bring your own keys, pay your provider directly, and pay us only for the work your agents do.
Billed yearly. Usage resets monthly.
For prototypes and side projects.
$0/ month
Free forever. No card required.
Includes
For small teams shipping their first production agent.
$49$39/ month
Billed yearly at $468.
Includes
For companies running agents across teams and systems.
$499$399/ month
Billed yearly at $4,788.
Includes
For security reviews, your own cloud and volume pricing.
Custom
Annual contract with volume pricing.
Includes
How billing works
01Step
One model call, tool call or adapter invocation inside an agent run. A typical support conversation takes 8 to 20 steps. Provider retries are free.
02Eval run
One test case run against a new version of your agent and scored against the current one. A 412-case dataset uses 412 eval runs.
03Tokens
Bring your own keys and pay OpenAI, Google, Mistral or any other provider directly. Prefer one bill? Clawsseum Credits are charged at provider list price.
Estimator
Pick a preset or drag the sliders. We price every plan at your usage and recommend the cheapest one that fits. Model tokens are not included: you pay your provider directly, or use Credits at list price.
People who write, test or deploy agents.
Model calls, tool calls and adapter invocations.
Test cases run against new versions of your agents.
Billed yearly. Usage resets monthly.
Recommended plan
Pro
$114/ month
Annual billing saves $120 a year on this plan.
Every plan at this usage
Estimates exclude taxes. Annual plans bill the platform fee yearly; overage is billed monthly in arrears.
Compare
Everything each plan includes, from adapters and eval limits to retention, SSO and support.
| Plan | FreeFree | Profrom $39 / mo | Teamfrom $399 / mo | EnterpriseCustom |
|---|---|---|---|---|
| Usage | ||||
| Builders included | 1 | 3 | 15 | Unlimited |
| Agents | 3 | Unlimited | Unlimited | Unlimited |
| Steps per monthOne model call, tool call or adapter invocation. Retries are free. | 10k | 250k | 3M | Custom |
| Eval runs per monthOne test case run against a new version and scored against the current one. | 500 | 5,000 | 50,000 | Custom |
| Model token markup | None | None | None | None |
| SDK and adapters | ||||
| TypeScript and Python SDKs | ||||
| Model adapters | All | All | All | All |
| Tool, data and channel adapters | 34 core | All standard | All + enterprise | All + custom |
| Bring your own MCP server or OpenAPI spec | ||||
| Scoped credentials per adapter | ||||
| Model fallback chains | ||||
| Evals | ||||
| Eval datasets | 3 | Unlimited | Unlimited | Unlimited |
| Build datasets from production traffic | ||||
| Model graders and deterministic checks | ||||
| Eval checks in CI | ||||
| Version-over-version scores | ||||
| Runtime | ||||
| Hosted runtime | Sandbox | |||
| Self-hosted or VPC runtime | ||||
| Approvals for side effects | Basic | Basic | Policies | Policies |
| Budgets and kill switches | ||||
| Trace retention | 7 days | 30 days | 90 days | Unlimited |
| OpenTelemetry export | ||||
| Security and support | ||||
| SSO (SAML, OIDC) | ||||
| SCIM and custom roles | ||||
| Audit log | ||||
| Data residency | US | US | US or EU | US, EU or your VPC |
| Uptime SLA | 99.9% | 99.95% | ||
| Support | Community | Shared Slack | Named engineer | |
FAQ
Anything we missed? Write to sales@clawsseum.com and a person will answer within a business day.
A step is one model call, tool call or adapter invocation inside an agent run. A typical support conversation is 8 to 20 steps. Retries caused by provider errors are free, and steps inside eval runs are not billed as steps.
An eval run is one test case run against a new version of your agent and scored against the current one. Running a 400-case dataset against one new version uses 400 eval runs.
No. Bring your own keys for Anthropic, OpenAI, Google, Mistral or any other provider and pay them directly. If you would rather not manage keys, Clawsseum Credits are billed at provider list price with no markup.
Nothing breaks. Overage is billed at the per-unit rates on this page and you get alerts at 80% and 100% of your included usage. Hard budget caps are available on every plan if you prefer a ceiling.
The SDK and local runtime work anywhere. A fully self-hosted control plane and runtime in your own VPC is part of Enterprise.
Yes. Early-access teams can run Team free for 21 days with production traffic. Ask for it when you request access.
Teams that join during early access keep their plan price for 24 months, even if list prices change.
Annual plans are billed up front at 20% off the monthly price. Included usage resets monthly either way.
Early access
Build and test your first agent on the Free plan. Early-access teams keep their price for 24 months.