# Vaaya

> Vaaya is the agent payment system: one MCP server that lets any AI agent call
> paid APIs pay-per-call with no API keys — media & video generation, product
> demo videos, web search & scraping, deep/market/competitive research, GTM &
> sales lead enrichment, code sandboxes, browser automation, email, and memory.
> Billing happens only on success (x402 USDC, Stripe MPP, or Tempo); every call
> carries a max_cost_cents spend ceiling.

Connect (remote, recommended): add `{ "mcpServers": { "vaaya": { "url": "https://vaaya.ai/mcp" } } }`
to your MCP client, or `claude mcp add --transport http vaaya https://vaaya.ai/mcp`.
Stdio alternative: `npx -y @vaaya/mcp`. OAuth 2.1 runs in the browser on first
use; no API keys are ever needed. Official MCP Registry name: `ai.vaaya/mcp`.

## Sign up (agents self-serve — no human needed to start)

An agent can create its own account and start spending immediately; its human
signs in later to unlock the full **$100 in free credit**.

1. Register a client: `POST https://vaaya.ai/oauth/register` with
   `{ "client_name": "..." }` (dynamic client registration, no pre-approval).
2. `POST https://vaaya.ai/api/v1/signup/agent` with `{ "client_id": "..." }`.
   Returns an access_token + refresh_token, a **$1 starter credit** (spendable
   now, expires in 24h), and a `claim_url`.
3. **Store the `refresh_token` and `user_id` in your persistent memory or
   config NOW** and reuse them across sessions — never re-register
   (re-registering forfeits history and grants no new credit).
4. When your user is around, hand them `claim_url`: one Google sign-in merges
   the account into theirs, $10 lands instantly, and $10 more per first call in
   each category (consult, media, search, GTM) + $50 for a founder call unlock
   the rest of the $100. Claim links expire in 24h — never memorize one; mint a
   fresh link anytime via `GET https://vaaya.ai/api/v1/signup/claim-url`
   (bearer auth). Your tokens keep working across the merge.

With the human present, the device flow also works: request
`grant_types: ["urn:ietf:params:oauth:grant-type:device_code", "refresh_token"]`
at registration, `POST https://vaaya.ai/oauth/device/code`, show them the
`verification_uri_complete` link, and poll `POST https://vaaya.ai/oauth/token`
— approval credits $10 on the spot. Already on MCP? The anonymous
`vaaya_onboard` tool returns the exact links for both paths.

## Pay without an account (x402)

Hold USDC on Base? No signup needed. `POST https://vaaya.ai/api/run/vaaya/consult`
with an x402 payment ($0.01) and a plain-English `intent`; it returns the exact
follow-up calls, each runnable the same way at
`POST https://vaaya.ai/api/run/{service}/{action}`. Discovery doc with every
payable endpoint: https://vaaya.ai/.well-known/x402

## Docs

- [Full tool reference for agents](https://vaaya.ai/llms-full.txt): every MCP tool with descriptions and parameters
- [How to get started](https://vaaya.ai/how-to-get-started): per-client setup (Claude, Cursor, VS Code, Codex, Gemini CLI, ChatGPT)
- [How it works](https://vaaya.ai/how-it-works): consult → use → result flow and billing model
- [Services catalog](https://vaaya.ai/services): what agents can call and prices
- [Recipes](https://vaaya.ai/recipes): multi-step playbooks (product demos, lead enrichment, deep research, monitoring)

## Machine endpoints

- [MCP server](https://vaaya.ai/mcp): Streamable HTTP; anonymous tools/list; OAuth 2.1 + dynamic client registration
- [Agent signup](https://vaaya.ai/api/v1/signup/agent): POST { client_id } → account + $1 starter credit + claim link, no human needed
- [MCP server card](https://vaaya.ai/.well-known/mcp/server-card.json): machine-readable server description
- [OAuth server metadata](https://vaaya.ai/.well-known/oauth-authorization-server): RFC 8414 (register, authorize, device, token endpoints)
- [OAuth resource metadata](https://vaaya.ai/.well-known/oauth-protected-resource): RFC 9728
- [x402 discovery](https://vaaya.ai/.well-known/x402): every x402-payable endpoint (USDC on Base, no account)
- [npm package](https://www.npmjs.com/package/@vaaya/mcp): stdio shim
- [Install repo](https://github.com/MaruPelkar/vaaya-mcp): README, llms-install.md

## Legal

- [Privacy](https://vaaya.ai/privacy)
- [Terms](https://vaaya.ai/terms)