Limits

Gabin is free during the preview. Limits keep the service fast for everyone and protect the data from bulk scraping. They apply per account: all your keys and OAuth tokens share them, and MCP tool calls count like API calls.

Free plan

Limit	Value	When exceeded
Rate	60 units per minute	`429 RATE_LIMITED` with `Retry-After`
Daily quota	5,000 units per UTC day	`429 QUOTA_EXCEEDED`
Concurrent requests	2 at a time per account	`429 RATE_LIMITED` with `Retry-After: 1`
Page size	100 items	—
Results reachable per query	10,000 (cursor pagination)	`400 PAGINATION_LIMIT`
Active API keys	5 per account	—

An IP address and its network also have a daily ceiling across all accounts (20,000 units per IP address, 60,000 per /24 IPv4 or /48 IPv6 network). You only reach it by spreading load over many accounts; the error is 429 QUOTA_EXCEEDED and its message says “this IP address” or “this network”. Higher limits are not available yet during the preview.

Request weights

A request costs 1 unit unless its endpoint says otherwise. The weight counts against both the rate and the quota.

Request	Units
Most endpoints	1
Advanced search: `POST /v1/search`, `POST /v1/search/count`	5
Map tile served from cache	0.1
Map tile computed from the database	0.5

The weight is listed in the description of each endpoint in the API reference. A request refused with RATE_LIMITED or BUSY because of concurrency costs nothing.

Response headers

Every /v1 response carries:

Header	Meaning
`X-RateLimit-Limit`	Units allowed per minute
`X-RateLimit-Remaining`	Units left in the current minute
`X-RateLimit-Reset`	Seconds until the minute window resets
`X-Quota-Remaining`	Units left today (UTC)
`X-Request-Id`	Identifier of the request, to quote if you report a problem
`Retry-After`	On `429` and `503`: seconds to wait before retrying

GET /v1/me returns the same counters as JSON, with today’s usage.

Handling limits in your code

Send requests one or two at a time per account. A third concurrent request is refused immediately.
On 429 or 503, wait Retry-After seconds, then retry. Back off exponentially if it happens again.
Do not retry QUOTA_EXCEEDED before the next UTC day.
Cache what you read: company and building data change at most daily.

async function gabin(path, { retries = 3 } = {}) {
  for (let attempt = 0; ; attempt++) {
    const res = await fetch(`https://api.gabin.ai/v1${path}`, {
      headers: { Authorization: `Bearer ${process.env.GABIN_API_KEY}` },
    });
    if (res.ok) return res.json();
    const body = await res.json().catch(() => null);
    const code = body?.error?.code;
    const retriable = res.status === 503 || (res.status === 429 && code !== "QUOTA_EXCEEDED");
    if (!retriable || attempt >= retries) throw new Error(`${res.status} ${code}: ${body?.error?.message}`);
    const wait = Number(res.headers.get("retry-after")) || 2 ** attempt;
    await new Promise((r) => setTimeout(r, wait * 1000));
  }
}

The MCP server already queues its own calls to stay within the concurrency limit.

⌘I

Get started

Guides

MCP server

Free plan

Request weights

Response headers

Handling limits in your code

​Free plan

​Request weights

​Response headers

​Handling limits in your code

Free plan

Request weights

Response headers

Handling limits in your code