Plans + limits

For interactive plan selection, see gnosyslabs.com/pricing. This page documents the limits in detail.

Plans

What's a "run"?

One closed propose-execute-validate-promote loop, regardless of how many iterations or specs it produces. A 5-iteration HP sweep that emits 25 specs total counts as one run.

The platform's definition (the row in the runs table that gets its own run_id) is what the quota counts. Each POST /v1/runs consumes one quota unit, returned at 202.

What counts against the LLM-tokens quota

The Starter and above plans bundle a token allowance for use with the LLM strategist + critic. Tokens consumed:

• Strategist round-N proposal: ~1-5K tokens depending on prior records context size.
• LLM critic per spec: ~500-2K tokens.

Typical Starter run with --llm-critic enabled, 5 iterations × 4 specs/round → ~50-200K tokens.

The token allowance does not apply if you bring your own LLM provider key (llm.api_key=...) — your provider bills you directly. We charge only for our pool tokens.

What happens when you hit the quota

POST /v1/runs returns HTTP 402 Payment Required with a body linking to the upgrade page:

{
  "detail": "monthly run quota exhausted on plan 'free' (5/5 used). Upgrade at https://gnosyslabs.com/pricing to continue."
}

In-flight runs continue to completion. Finding queries (GET /v1/findings) and historical run reads (GET /v1/runs/{id}) are unrestricted — only the create path enforces.

Burst rate limits

Separate from the monthly quota. Protects the platform from runaway automation. Per-tenant per-minute window. Returns HTTP 429 with Retry-After: 60. The SDK respects Retry-After automatically and retries up to max_retries (default 3).

The limit applies only to POST /v1/runs. All read endpoints are unrestricted; query as much as you like.

Plan changes

Self-serve from /dashboard/billing (Phase F — coming soon). Until then, contact sales@gnosyslabs.com to switch.

Quota counters reset on the first of every month UTC. When you upgrade mid-month, the new plan's limits apply immediately and your counter is preserved (so you can keep running where you left off).

Enterprise / self-hosted

The platform engine is closed-source SaaS by default. The Enterprise tier offers self-hosted deployment with:

• Engine licensed for on-prem deployment
• Dedicated GPU pool (your choice of cloud)
• SAML + SCIM
• SOC2 evidence + DPA
• Custom domain (e.g. gnosys.acme.com)
• Dedicated CSM

Talk to sales.

FAQs

Can I bring my own LLM key? Yes, on Team and above. Pass llm.api_key="sk-..." in the run-create payload. The platform forwards calls through your provider; tokens don't count against our pool.

What happens to my runs if I downgrade? Existing runs and findings stay queryable forever. Historical data is part of the audit trail; we don't delete it on plan changes.

Can I get a refund if I exhaust my quota in the first day? Standard SaaS terms. Reach out for credit-on-account if you have a specific issue and we'll work with you.

Is there a usage-based pricing tier? Coming Q3 2026. Pay per spec executed, no monthly minimum. Email sales to be notified.

Found a typo? Tell us.