vaaya / about
Vaaya connects AI agents to paid services and handles the provider accounts, credentials, and billing. Provider keys stay with Vaaya. Learn who builds the product, how requests are handled, and how you can control spending.
An agent describes an outcome. Vaaya routes it to a suitable provider, validates the request, shows the price ceiling, executes the call, and returns the result. Search, data, LLMs, media, scraping, compute, and storage all sit behind the same connection. The gateway architecture is described in the white paper.
Calls are metered individually and failed calls are not billed. See the public capability catalog, pricing, and documentation.
Vaaya is built and maintained by Vaaya Inc. The MCP connector and integration work are visible through the vaaya-ai GitHub organization, where developers can inspect releases, open issues, and follow the project.
We want adding a service to an agent to be as simple as asking it to do the work.
Your agent uses a Vaaya connection to call services. You can limit its spending and revoke its access without changing provider credentials.
A call cannot run without sending its parameters to the provider that serves it. Here is what Vaaya stores and forwards, per data type. Across every row: your data is never used to train public or shared foundation models, never sold, and never licensed to third parties (Privacy Policy).
| Data | Stored by Vaaya | Sent to providers | Retention | Your control |
|---|---|---|---|---|
| Call parameters & results | Yes, in your transaction history, so results can be replayed instead of re-billed | Parameters go to the one provider serving the call | Kept with the account | Visible in your history; deleted with your account |
| Consult conversations | Yes, so routing improves and you can review what was asked | No, routing happens inside Vaaya | Kept with the account | Visible in your history; deleted with your account |
| LLM router requests | No. Per-call metadata only: model, tokens, price, calling client, a hash of the system prompt, a scrubbed 200-character prefix of the first message | The full request goes to the one model provider serving it | Metadata kept with the account. Full bodies are captured only for accounts flagged for abuse, and deleted after 7 days | Deleted with your account |
| Billing metadata (price, provider, outcome) | Yes, this is the ledger | No | Kept for accounting and tax obligations | Exportable from your history |
| Files you upload | Yes, in your file library, billed once | Only when a call you make uses them | Until you delete them | Delete any time |
| Company memory (brain) | Yes, scoped to your workspace | No | Until you clear it | Edit or clear from the app |
| GitHub profile data | The computed score and its inputs, public data only | No | Refreshed at most daily; kept while connected | Disconnect any time; the read:user grant is revocable at GitHub |
| Provider credentials | Held server-side, encrypted | Injected per validated call, never returned in a response | Rotated by Vaaya | Not applicable, they are Vaaya's accounts |
Email hello@vaaya.ai, or open an issue in the public MCP repository. Legal and privacy contacts are listed in our Privacy Policy and Terms.