Rate limits - Lanesurf Agent API

Each API key can make 120 requests per minute (about 2 a second) on the authenticated /v1 routes and on /mcp. Over a limit you get 429 rate_limited with a Retry-After header in seconds.

The limits

What	Limit	Counted per
Authenticated `/v1` routes and `/mcp`	120 requests per minute	API key
Signup starts (`POST /v1/signup`, `start_signup`)	5 per hour and 10 per day	Email address
Signup starts	10 per day	MC or DOT number
Signup starts	200 per hour	Client IP
Signup routes (`/v1/signup` start and verify, the MCP signup tools)	300 per 10 minutes	Client IP
Wrong API keys	100 per 10 minutes	Client IP (refused before the key is checked)
OAuth sign-in page steps (password, key and code attempts)	60 per 10 minutes	Client IP
Signup codes	5 wrong codes end the signup; a code expires after 10 minutes	Signup
Request body	1 MB	Request

The per-IP limits are high on purpose: every user of one AI platform (ChatGPT, Claude, Grok) shares a few of its IP addresses. The limits that matter are per key, per email and per MC/DOT number. The per-minute and per-10-minute limits are counted on each Lanesurf server, so treat them as the most you should send, not a target.

What a 429 looks like

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

{ "error": "too many requests for this API key; slow down", "code": "rate_limited", "request_id": "req_…" }

Signup 429s also carry next_step. Over MCP, a limited tool call returns the same JSON as an error result, with retry_after_seconds.

How to back off

On 429, wait the Retry-After seconds (or retry_after_seconds), then retry once.
Don’t retry in a tight loop. A second 429 means wait again, longer.
Poll slowly. A load’s offers change over minutes, not seconds: poll GET /v1/loads/{id} or GET /v1/carrier/bids every few minutes. Stop polling a load that is covered, inactive or deleted, and slow down on stalled.
Spread bulk work (many trucks or lanes) out instead of sending it all at once.

Retrying a POST /v1/loads after a 429 is safe with the same Idempotency-Key. See Idempotency and retries.

Every limit answer uses the code rate_limited. See Errors for every other code.