platform pricing
Every limit published. Nothing quoted by email except Enterprise.
The numbers below come from the single file the API enforces. Free gives you a working sandbox key with 10K requests a month and no card. Startup at $99 is the first plan that can issue live keys. Growth at $499 multiplies the quota tenfold. Enterprise is where you bring your own provider account.
the four plans
Pick the one whose ceiling you will hit last.
Every plan carries the whole platform — all endpoints, all SDKs, AI enrichment, webhooks, the MCP server. What changes between them is volume, rate, and whether keys can go live.
Free
Free
Build the whole integration against the sandbox. No card, no approval queue.
- 100 connected users
- 10K API requests / month
- 1K AI requests / month
- 10 requests / second sustained
- 1 webhook endpoint
- 3 team members
- Sandbox environment only
Startup
Most teams start here$99/ month
The first plan that can issue live keys. Where most teams go to production.
- 5,000 connected users
- 500K API requests / month
- 50K AI requests / month
- 50 requests / second sustained
- 5 webhook endpoints
- 10 team members
- Live keys for real institutions
Growth
$499/ month
Ten times the quota and four times the rate limit, for products with traction.
- 50,000 connected users
- 5M API requests / month
- 500K AI requests / month
- 200 requests / second sustained
- 20 webhook endpoints
- 25 team members
- Live keys for real institutions
Enterprise
Custom
Your own provider accounts, your own deployment, quotas set by contract.
- Connected users by agreement
- Request volume by agreement
- AI volume by agreement
- 1000 requests / second sustained
- 100 webhook endpoints
- 250 team members
- Live keys for real institutions
metering
What counts as a request.
Billing arguments come from vague metering. Here is the whole rule, in the order the API applies it.
One authenticated /v1 call, success or failure
A request that resolves to your application is metered even if it then fails validation — a400for a malformed body still consumed the work. A call with an invalid or revoked key is not counted, because it never reaches an application.AI endpoints count against two counters
A call to any/v1/ai/*endpoint increments both API requests and AI requests. That is why the AI quota is a tenth of the API quota on every plan: it is a sub-budget of the same traffic, not a second bill.Sandbox traffic is real traffic
Test keys and live keys meter into the same monthly counters. It costs us the same compute either way, and a quota that only counts production would make your pre-launch numbers meaningless.Counters are per application, per UTC month
Usage is keyed by application, month and product, and every served request also writes a log row with its path, status, latency and request id. The usage screen and the counters are reading the same data, so they cannot disagree.Rate limit and quota are different ceilings
The monthly quota is a volume budget. The rate limit is a token bucket per key at your plan's sustained rate with a burst of 3×, and every response carriesX-RateLimit-Limit,-Remainingand-Reset. Hitting either returns429.Webhooks and retries are free
Outbound deliveries never consume quota, including the retry schedule. You are not billed for us telling you something changed.
side by side
Every entitlement, every plan.
These values mirror the entitlements file the API reads on every request. If a number here is wrong, the API is wrong with it.
| Entitlement | Free | Startup | Growth | Enterprise |
|---|---|---|---|---|
| Price per month | Free | $99 | $499 | Custom |
| Connected users | 100 | 5,000 | 50,000 | By agreement |
| API requests per month | 10K | 500K | 5M | Negotiated |
| AI enrichment requests per month | 1K | 50K | 500K | Negotiated |
| Requests per second(burst 3×) | 10 | 50 | 200 | 1000 |
| Webhook endpoints | 1 | 5 | 20 | 100 |
| Team members | 3 | 10 | 25 | 250 |
| Sandbox environment | Included | Included | Included | Included |
| Live keys for real institutions | Not included | Included | Included | Included |
| All endpoints, including AI enrichment | Included | Included | Included | Included |
| Every SDK, MCP server, OpenAPI and webhooks | Included | Included | Included | Included |
| Bring your own provider account(onboarded with us) | Not included | Not included | Not included | Included |
| Private deployment | Not included | Not included | Not included | Included |
in every plan
The whole platform, from the free tier up.
Nothing below is an add-on, an upsell, or a feature flag we turn on after a call.
| Included | Detail |
|---|---|
| Endpoints | 22 core plus 9 AI enrichment — 31 under /v1, all POST with a JSON body |
| Institution catalogue | 938 institutions across 61 countries, routed to the best available provider |
| Client libraries | 11 SDKs, from the browser Link embed to .NET |
| Sandbox | A simulated institution network with 24 months of deterministic history per item, seeded so the same connection always returns the same data. Self-serve end to end: sign up, verify your email, create an application, issue a test key — no approval queue and no sales call anywhere in that path |
| Webhooks | HMAC-SHA256 signed with a timestamp, retried on a backoff schedule, with deliveries recorded per endpoint |
| AI agent surface | MCP server over Streamable HTTP, OpenAPI 3.1, llms.txt and llms-full.txt, plus the two .well-known descriptors |
| Console | Applications, keys, webhooks, usage, connected users, request logs, error explorer, team and audit log |
| Manual and on-chain paths | CSV, OFX and QFX import plus public-address crypto tracking — both work in live mode with no provider agreement at all |
going live
What live access actually requires.
Two conditions, and the second one is the part every vendor page skips.
01
A plan that can issue live keys
Startup or above. In the dashboard you enable live access on the application, then create a live secret key — shown once. Nothing else about your integration changes: same endpoints, same request bodies, same response shapes, same cursors.
02
A provider for the institutions you need
Reading someone's bank account is a regulated activity, so every upstream provider requires an agreement signed by a legal entity. Bring your own on Enterprise, or use the credentials we configure. Institutions we cannot reach are hidden from search rather than faked.
before you commit
The questions a finance lead asks second.
Keep reading
- SolutionsWhat the platform looks like for AI agents, budgeting apps, lenders, accounting tools and wealth products — with the endpoints each one calls.
- EnterpriseRun connections on your own Plaid, Teller or Enable Banking account, with per-org credential isolation, audit logs and a private deployment.
- API documentationThe full reference, the interactive explorer, and a playground that runs real sandbox calls.
Start in the sandbox. Decide about us afterwards.
A Free-plan key, no card and no approval queue — then 10K requests a month to find out whether the shapes fit your product.
Enterprise pricing depends on volume, which provider accounts you bring, and whether you need a private deployment. There is no list price for it because there is no honest one.