Discovery

Discovery documents are free and generated from the route registry + PRICING at request time — zero price literals exist outside the single price source.

GET /openapi.json (free)

OpenAPI 3.1 specification covering the full route inventory, with the free/paid split and x-tollbridge price annotations on paid operations.

curl https://api.tollbridge.app/openapi.json

GET /.well-known/x402.json (free)

x402 payment discovery: for every paid route, the scheme (exact), network (eip155:8453), asset (USDC Base mint), amount (integer atomic USDC) and payTo — the exact requirements a client needs to pre-sign.

curl https://api.tollbridge.app/.well-known/x402.json

GET /.well-known/mcp.json (free)

MCP server + tool discovery: the Streamable HTTP endpoint and the tool catalog with per-tool prices.

curl https://api.tollbridge.app/.well-known/mcp.json

GET /.well-known/x402/mcp.json (free)

The same MCP discovery document, mounted at the path the x402/agent convention names — an agent that found the payment rail first can read the tool list without a second lookup.

curl https://api.tollbridge.app/.well-known/x402/mcp.json

GET /.well-known/ai.txt (free)

Agent discovery pointer: a plain-text list of the documents an agent should read first — llms.txt, llms-full.txt, /.well-known/x402, /.well-known/mcp.json, openapi.json — all on the live origin, with the per-decision price.

curl https://api.tollbridge.app/.well-known/ai.txt

GET /.well-known/agent.json (free)

Agent identity card: name, description, version, docs URL, payment rail and the endpoint list with free/paid flags.

curl https://api.tollbridge.app/.well-known/agent.json

GET /robots.txt (free)

Crawler policy for the live origin: Allow: / with /v1/admin/ disallowed. The reserved product hostnames are owned but not attached — see DOMAINS.md.

curl https://api.tollbridge.app/robots.txt

GET /llms.txt (free)

llmstxt.org v2 specification: the first forty lines are the agent loop alone — pay, poll, feed, next — with the tool names, the per-decision price on Base, and the reserved-host footnote. The reference sections follow.

curl https://api.tollbridge.app/llms.txt

GET /llms-full.txt (free)

The complete agent recipe and endpoint reference — the same document as llms.txt, served under the conventional "full" name crawlers probe.

curl https://api.tollbridge.app/llms-full.txt

TypeScript

Discovery documents are plain HTTP JSON — consume them with any client. The SDK wraps the functional surfaces, not the discovery files:

import { Tollbridge } from '@tollbridge/sdk-typescript';

const toll = new Tollbridge({ baseUrl: 'https://api.tollbridge.app' });

// The SDK reads the price list through the free pricing route:
const { data } = await toll.pricing();

Python

import httpx

# Reference snippets — official Python SDK in progress.
base = "https://api.tollbridge.app"
openapi = httpx.get(f"{base}/openapi.json").json()
x402 = httpx.get(f"{base}/.well-known/x402.json").json()
mcp = httpx.get(f"{base}/.well-known/mcp.json").json()
agent = httpx.get(f"{base}/.well-known/agent.json").json()
llms = httpx.get(f"{base}/llms.txt").text

MCP

The MCP server is itself discoverable — .well-known/mcp.json publishes the Streamable HTTP URL; over a live connection the tool catalog is one tools/list away:

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/list",
  "params": {}
}

{
  "tools": [
    "tollbridge_market_state",
    "tollbridge_liquidation_risk",
    "tollbridge_funding_opportunities",
    "tollbridge_wallet_intelligence",
    "tollbridge_transaction_risk",
    "tollbridge_token_risk",
    "tollbridge_decision"
  ]
}