# GeekVault > Pay-per-call API catalog for AI agents, over the x402 protocol (HTTP 402 + USDC on Base). No API keys, no subscription, no account — sign a payment, get the response. GeekVault exposes utilities behind `api.geekvault.dev`. Each call costs between $0.002 and $0.015 USDC, settled on Base mainnet (`eip155:8453`) via x402. Machine-readable payment manifest: https://api.geekvault.dev/.well-known/x402 ## Categories - [Dev Tools](https://geekvault.dev/#dev-tools) - [Image Tools](https://geekvault.dev/#img-tools) - [Docs Tools](https://geekvault.dev/#docs-tools) - [Network Tools](https://geekvault.dev/#net-tools) - [Web Tools](https://geekvault.dev/#web-tools) - [AI Tools](https://geekvault.dev/#ai-tools) ## How it works - [How x402 works in 3 steps](https://geekvault.dev/#how-it-works) - [Integration guide (curl + @x402/fetch + MCP)](https://geekvault.dev/#integration) ## MCP GeekVault is also available as a remote MCP server at https://mcp.geekvault.dev/mcp (registered as `dev.geekvault.mcp/gv-mcp` in the official MCP registry, registry.modelcontextprotocol.io). Same x402 payment flow, exposed as MCP tools via Streamable HTTP — tools are discovered via `tools/list`, not statically enumerated here. ## Payments Payments are idempotent by `payment_id` (x402 payment-identifier extension). Retrying the same signed payment is safe and won't be charged twice. ## Discovery - [OpenAPI spec](https://api.geekvault.dev/openapi.json): full parameter/format/response schema for every endpoint, generated directly from the server source code. - [x402 payment manifest](https://api.geekvault.dev/.well-known/x402): every endpoint with method, price and network, for automated discovery by x402-aware agents and catalogs. - [API catalog (RFC 9727)](https://api.geekvault.dev/.well-known/api-catalog): standard discovery pointer to the OpenAPI spec. - [Full text version](https://geekvault.dev/llms-full.txt): everything on this page plus the complete endpoint catalog, in plain text. - [Health check](https://api.geekvault.dev/health): free, no payment required. ## Latency Real p50 latency (measured, not estimated) is in the `p50LatencyMs` field of every resource in the x402 manifest. Most calls respond in well under 1 second. The slowest: - `POST /ai-tools/v1/transcribe`: ~13.4s - `GET /net-tools/v1/ping`: ~3.0s - `GET /web-tools/v1/pdf-export`: ~2.9s - `GET /web-tools/v1/screenshot`: ~2.3s - `GET /net-tools/v1/traceroute`: ~2.0s