Rate limits & quota
Frontwind metering today is plan-based, not per-request.
Plans
| Plan | Search quota | Notes |
|---|---|---|
| Free | 2 lifetime searches | Each call to a quota-consuming agent (e.g. Competitive Landscaping, Target Prioritization, Patent Mining) decrements the lifetime counter. CRUD agents like alerts do not consume quota. |
| Pro | Unlimited | No per-request limits today. Concurrency is implicit — each call runs through the existing pipeline scheduler. |
Concurrency
A single user can have one active long-running agent run at a time (Competitive Landscaping / Target Prioritization / Patent Mining). Submitting a second while one is still running returns a 409 from the underlying pipeline — wait for the first to finish.
HTTP timeouts
Cloud Run caps each request at 60 minutes. Set your client timeout to at least 3600s. For shorter agents (Patent Structures, Drug Patent FTO) typical p95 latency is well under 10 minutes.
What we plan to add
- Per-key request-rate limits with
X-RateLimit-*headers. - Async job mode with a
run_idrecovery endpoint, so 60+ minute pipelines survive disconnects. - Webhooks for long pipelines.