Usage Limits

Overview

Limits follow the organization of the resource — your personal workspace (Free or Pro) or a team organization (Enterprise on your cloud). Each workspace has a per-minute and per-day API request limit, plus caps on projects, agents, apps, MCP connectors, and agent invocations.

In product terms, one invocation is one call to run an agent (agent.invoke()or POST /agents/{id}/invoke). Many other authenticated platform API calls (deploy, identity / user management, jobs, sessions) also share the same daily / minute meters.

Where to find this page

• Docs menu — Navbar → Docs → Usage Limits
• Console dashboard — Daily API usage card → plan badge, % used, or “View what counts →”
• Direct URL — /docs/usage-limits

What counts at a glance

Invocations (the main consumer)

Each accepted attempt to run an agent counts as one requesttoward your limits, regardless of response length or token usage.

• Python Agent SDK — client.agent("…").invoke(…)
• Node Agent SDK — same invoke / stream methods over HTTP
• App SDK — app-side agent invoke helpers (Python or Node)
• Playground — Messages you send in the console playground
• HTTP — curl or custom clients calling POST /api/v1/agents/{agent_id}/invoke
• Streaming invoke — Stream endpoints still count as one request when the call is accepted under rate limiting

See also: Invoke an Agent · SDK Reference

Deploy & manage

Deploying and managing agents through the SDK / platform API (with your API key) also consumes the same request quota — for example:

• SDK / HTTP deploy of a new agent (Python or Node)
• Update or delete agent via SDK-authenticated routes
• List or fetch agent details via those same authenticated API routes

Console UI flows that only open pages do not burn quota by themselves; actions that hit the rate-limited agent APIs do.

Other API calls on the same meter

These surfaces share the developer rate limiter when authenticated with an API key / SDK token. Each accepted call counts as one request:

• User management / identity — create users, list users, delete users, login, refresh, logout, MFA, SSO, password flows under /api/v1/identity/…
• End-user APIs — end-user management routes authenticated as the developer
• Jobs — enqueue / job APIs
• Sessions — create, list, get, TTL, context get/set/delete
• Node SDK — identical to Python; every SDK method that calls these HTTP routes counts

What does not finish your limit

• Reading documentation or this page
• Opening the console dashboard, projects, API keys, or settings without calling rate-limited APIs
• Google / Cognito sign-in for the developer console itself
• Static asset loads (images, JS bundles)

Plan limits

Caps apply to the workspace you are in. Personal orgs start on Free; Hikigai admin can grant Pro. Creating a team organization is Enterprise and requires your own cloud to deploy. Existing resources keep working if you are over a cap; new creates are blocked.

Console dashboard

The Dashboard follows the org switcher: personal workspace KPIs, or this organization's tiles. Owners and admins of a team org also see all-members Daily API usage above their own.

• Used / limit — requests used today vs your plan's daily cap
• Remaining — how many requests you have left in the daily window
• Per minute — current minute usage vs the RPM cap
• Plan badge — Free / Pro / Enterprise (clickable → this page)

When you hit the limit

Exceeding the minute or day quota returns HTTP 429 with rate-limit headers (X-RateLimit-*, Retry-After). Wait for the window to reset, or move to a higher plan.

Track remaining quota anytime on the console dashboard.