Skip to Content
Limits & Billing

Limits & billing

Operational caps the agent enforces at runtime, plus how credits and plans work. Design your tools and conversations against these numbers.

Runtime caps

Per-conversation and per-call limits enforced by the chat engine.

CapValueWhat happens on overrun
Tool result size8 KBResult is truncated server-side (...truncated suffix) and the truncated value is what the agent sees.
Route timeout30sThe server-to-server fetch is aborted and the tool call returns a timeout error.
Steps per conversation20The agent stops generating further steps once the limit is reached (graceful stop, no error).
Tokens per conversation80,000The conversation is hard-stopped with a token_cap error. Start a new conversation to continue.
Idempotency key TTL24 hoursDuplicate execute-result calls with the same key return the original cached response instead of re-executing.

Keep route/server-action handlers fast and their return values small — design for the 8 KB / 30s caps above, not just “concise.”

Rate limits

Enforced per endpoint via a sliding window. Exceeding a limit returns 429 Too Many Requests with a Retry-After header.

EndpointLimitKeyed by
POST /api/v1/chat20 req/minproject + end-user
POST /api/v1/chat (project ceiling)120 req/minproject
POST /api/v1/execute-result60 req/minproject
POST /api/v1/describe30 req/minproject
Auth endpoints (sign-in, OTP verify)20 req/minIP
Sign-up5 attempts / 15 minIP

Models

The hosted chat engine currently runs on a single fixed model — claude-sonnet-4-6 — for every project and plan. Per-project or per-conversation model selection isn’t available yet.

The model-picker widget on the marketing homepage is illustrative of where this is headed, not a live setting — there’s no way to switch to Haiku or Opus today.

Credits & plans

Every plan is billed in credits, consumed per action regardless of model or token count:

ActionCredits
Conversation start2
Message1
Tool call3

Plans differ in included credits and whether you can go over:

PlanIncluded creditsPeriodOverage
Free50030 daysHard-capped — chat endpoint returns an error until reset, no extra billing.
Starter1,50030 daysHard-capped, same as Free.
Plus4,00030 daysAllowed, $10 per 1,000 additional credits. Supports bring-your-own API key for unlimited usage.
EnterpriseCustom30 daysCustom — talk to sales .

The credit period is a rolling 30-day window — it resets 30 days after the window last started, not on a fixed calendar date.

Full CLI reference: CLI Reference. Tool result/runtime behavior: Tool Files.

Last updated on