Vaaya MCP — payments for any agent

MCP server for AI agent payments. An agent in Cursor, Claude Code, Claude

Desktop, Codex, or any MCP host can call paid APIs (search, scraping, contact

data, public records, compute, media, research) from one balance, with a spend

ceiling it states on every call, and no vendor keys anywhere in its

environment. Billed only on success (x402 USDC on Base, Stripe MPP, or Tempo);

a refused or failed call costs nothing.

npx @vaaya/mcp install            # Cursor, Claude Code, Claude Desktop, Codex
claude mcp add --transport http vaaya https://vaaya.ai/mcp

First paid call: `consult` returns a priced plan; `use({ service, action,

params, max_cost_cents })` runs one line of it. If the quoted price is above

`max_cost_cents`, the call is refused before it reaches the provider and the

reason comes back:

{ "error": "over_max", "message": "price is 5¢ but max_cost_cents was 1", "price_cents": 5 }

Runnable examples: `examples/pay-for-an-api` (one

call allowed, one refused), `examples/cursor`,

`examples/claude`. Guides:

give an agent a budget,

Cursor setup,

Claude setup,

what is an AI agent wallet.

What your agent can do through Vaaya

Install

One command (recommended)

npx @vaaya/mcp install

(`npx vaaya-cli install` runs the same installer.)

Sets up the Vaaya MCP server for every agent it detects on your machine —

Claude Code, Claude Desktop, Cursor, Codex — plus the agent skill.

Idempotent: re-run any time to update. Then restart your agent; the first

Vaaya call opens a short browser approval (sign-up happens right there).

Remote (Streamable HTTP + OAuth)

{ "mcpServers": { "vaaya": { "url": "https://vaaya.ai/mcp" } } }

Claude Code:

claude mcp add --transport http vaaya https://vaaya.ai/mcp

OpenClaw:

openclaw mcp add vaaya --url https://vaaya.ai/mcp --transport streamable-http --auth oauth
openclaw mcp login vaaya

Hermes (`~/.hermes/config.yaml`):

mcp_servers:
  vaaya:
    url: "https://vaaya.ai/mcp"
    auth: oauth

On first use your client runs the OAuth flow in the browser; approve and you're

connected. Revoke anytime at vaaya.ai/connected-apps.

npm stdio shim (for stdio-only clients: opencode, …)

{ "mcpServers": { "vaaya": { "command": "npx", "args": ["-y", "@vaaya/mcp"] } } }

The shim (`@vaaya/mcp`) opens the

OAuth flow on first call, stores a refresh token locally (`0o600`), and proxies

the live tool list from the backend — new server-side tools appear without a

shim upgrade. It also ships an agent skill so your client routes every

capability gap through `consult` automatically.

Agent skill (works across 70+ agents)

npx skills add vaaya-ai/vaaya-mcp

Installs the `vaaya` skill for Claude Code, Codex,

Cursor, Gemini CLI, Copilot, and any other client that supports the open

Agent Skills standard. The skill bootstraps the

Vaaya MCP server if it's missing and keeps the install updated.

As a plugin

• Claude Code: `/plugin marketplace add vaaya-ai/vaaya-mcp` then `/plugin install vaaya@vaaya` — bundles the remote MCP server, the `vaaya` skill, and the SessionStart reminder hook
• Codex: `codex plugin marketplace add vaaya-ai/vaaya-mcp` then install `vaaya` from the Plugins panel
• Gemini CLI: `gemini extensions install https://github.com/vaaya-ai/vaaya-mcp`

How it works

1. `consult(intent)` — describe any goal in plain English ("generate a hero

video", "find 50 heads of RevOps with verified emails", "watch my

competitor's pricing page"). It converses, clarifies, and returns the exact

call(s) to run.

2. `use(service, action, params, max_cost_cents)` — execute the call.

Billed on success; `max_cost_cents` is a hard spend ceiling. Long jobs

return `{ async: true, job_id }` — poll with `result(job_id)`.

3. `session` / `close` — per-second-billed code sandbox sessions.

Plus a full GTM suite (`gtm_leads_find`, `gtm_lead_enrich`, `gtm_signal_create`,

`gtm_segments`, `gtm_message`, `gtm_replies`, …) and scheduled monitoring

workers (`worker_create`, `worker_findings`, …) — the tool list is proxied live from the backend.

Example prompts

• "Generate a hero video for our landing page in 16:9."
• "Record my product and turn it into a narrated demo video."
• "Find 50 heads of RevOps at Series-B SaaS companies with verified emails."
• "Watch my competitor's pricing page and tell me when it changes."
• "Do deep research on the agentic payments market with cited sources."
• "Scrape these 200 product pages into structured JSON."
• "Run this benchmark in a GPU sandbox."

CLI

The `vaaya-cli` CLI manages everything from

the terminal:

npx vaaya-cli install        # set up the MCP server for every detected agent
npx vaaya-cli status         # connection state + live tool count
npx vaaya-cli consult "..."  # ask Vaaya how it would do something (returns the plan)
npx vaaya-cli reauthorize    # re-run the browser auth flow
npx vaaya-cli logout         # disconnect & delete local credentials

Security & billing

• Your agent never sees vendor API keys or upstream URLs — Vaaya proxies every

call server-side.

• OAuth 2.1 (PKCE, dynamic client registration); anonymous `tools/list` for

discovery; revocable grants.

• Pay-per-call on x402 (USDC on Base), Stripe MPP/SPT, or Tempo. Failed calls

are never charged. Every call carries a `max_cost_cents` guard.

Links

• Website & docs: https://vaaya.ai
• Agent-readable docs (llms.txt): https://vaaya.ai/llms.txt · full tool reference: https://vaaya.ai/llms-full.txt
• npm: https://www.npmjs.com/package/@vaaya/mcp · https://www.npmjs.com/package/vaaya-cli
• Support: support@vaaya.ai