Rate limits

Lanes Forms tells you where you stand instead of making you find out by being refused. Every response on a throttled path carries its quota policy, and where the server has actually counted, the live remaining quota too.

The headers

RateLimit-Policy: "submissions";q=10;w=60
RateLimit: "submissions";r=7;t=42
RateLimit-Limit: 10
RateLimit-Remaining: 7
RateLimit-Reset: 42

RateLimit-Policy and RateLimit are the fields from the IETF draft-ietf-httpapi-ratelimit-headers specification. In RateLimit-Policy, q is the quota and w is the window in seconds. In RateLimit, r is the quota you have left and t is the seconds until the window resets. The quoted name ties the two together, so a response can carry more than one policy without ambiguity.

RateLimit-Limit, RateLimit-Remaining, and RateLimit-Reset are the older spelling of the same three numbers. The draft supersedes them, but plenty of clients only understand those, so both are sent.

On a 429

Every 429 carries Retry-After with a number of seconds:

HTTP/1.1 429 Too Many Requests
Retry-After: 60
Content-Type: application/json

{"error": {"code": "submission_rate_limited", "message": "...", "docs_url": "..."}}

Wait that long, then retry. If a response carries both Retry-After and RateLimit, Retry-After wins.

For a limit measured over 24 hours, Retry-After is capped at one hour rather than reporting the full window. Telling a caller to sleep for a day is not actionable, and retrying after an hour costs one refused request and gets you a fresh number.

The documented limits

Exceeding one of these returns a 429 with a specific code: submission_rate_limited, unclaimed_submission_cap, form_creation_rate_limited, domain_rate_limited, claim_email_rate_limited, pending_claim_exists, or api_key_limit_reached. Each is listed with its fix in Errors.

Anonymous form creation also has a service-wide ceiling that answers 503 anonymous_provisioning_paused rather than 429. It is a temporary pause, not a per-caller throttle, so it is not something backing off harder will clear.

Staying under them

• Read RateLimit-Remaining and slow down before it reaches zero, rather than sprinting into a 429.
• Use a workspace API key. It raises the MCP transport limit from 60 to 300 requests per minute and provisions forms already claimed, skipping the anonymous-path gates.
• Batch nothing that does not need batching. The per-IP submission limit is per form, so spreading real traffic across forms is fine; hammering one is not.
• Quote x-request-id from the response if you believe a limit is wrong.