Quickstart

Connect to your AI assistant

Install the MCP server and your AI client gets 6 tools to search the index.

npm install -g agent-discovery-mcp-server

// Add to ~/Library/Application Support/Claude/claude_desktop_config.json
{
  "mcpServers": {
    "montexi": {
      "command": "npx",
      "args": ["-y", "agent-discovery-mcp-server"]
    }
  }
}

Restart Claude Desktop. The tools appear automatically.

claude mcp add montexi -- npx -y agent-discovery-mcp-server

// Add to .cursor/mcp.json
{
  "mcpServers": {
    "montexi": {
      "command": "npx",
      "args": ["-y", "agent-discovery-mcp-server"]
    }
  }
}

// Add to settings.json
{
  "mcp": {
    "servers": {
      "montexi": {
        "command": "npx",
        "args": ["-y", "agent-discovery-mcp-server"]
      }
    }
  }
}

Then ask your AI:
"Find me free translation APIs"
"What MCP tools can search databases?"
"Show me the index stats"
"Check if api.example.com is agent-ready"

Use in your code

Find services in 3 lines.

npm install agent-discovery-sdk

import { AgentDiscovery } from 'agent-discovery-sdk'
const discovery = new AgentDiscovery()
const results = await discovery.search('translation')

More examples

// Find free services only
const free = await discovery.search('search', { access: 'free' })

// Find MCP servers with tools
const mcp = await discovery.search('*', {
  sourceType: 'mcp-server-card',
  hasTools: true
})

// Search MCP tools directly
const tools = await discovery.tools('search')

// Async iteration
for await (const svc of discovery.discover('geocoding')) {
  console.log(svc.endpoint, svc.readiness_score)
}

Get discovered in 2 minutes

Make your service findable by any AI agent.

Submit your service

Can't wait for the crawler? Submit your endpoint and we'll index it now.

Or get discovered automatically

Already have an OpenAPI spec at /openapi.json? You're already discoverable. The crawler finds you automatically.

// Serve at /.well-known/agent-card.json
{
  "name": "My Service",
  "url": "https://api.example.com",
  "skills": [{
    "id": "search",
    "name": "Search"
  }]
}

/.well-known/agent.json and /llms.txt are also supported discovery protocols. The crawler checks all paths on every domain.

How agent-ready is your service? Check Your Score

Deep Dive

What We Index Protocol breakdown with live counts

The crawler probes thousands of domains across multiple discovery protocols.

A2A Agent Cards

—

/.well-known/agent-card.json

MCP Servers

—

/.well-known/mcp/server-card.json

OpenAPI Services

—

/openapi.json, /swagger.json

LLM-Accessible

—

/llms.txt

MCP Tools

—

Individual tools via JSON-RPC handshake

How Services Are Accessed Access classification breakdown

Every alive service is probed and classified by its access requirements.

Free

—

No authentication needed

API Key

—

Requires an API key or bearer token

OAuth

—

OAuth2 or OpenID Connect

Payment-Gated

—

Requires payment (x402, Stripe, etc.)

The Economic Layer x402 services with live pricing

Services that accept real-time micropayments via the x402 protocol. No accounts, no billing — pay per request with BSV.

Loading x402 services...

x402 extends HTTP with settlement-gated access. Agents pay per request using BSV micropayments — no API keys, no subscriptions, no billing infrastructure. Learn more about x402

Agent Readiness Scoring system explained

Every indexed service receives a readiness score from 0 to 100, measuring how easily an AI agent can discover and use it.

Discovery

30 pts — Has agent cards, well-known files, or specs

Accessibility

25 pts — Free access scores highest

Responsiveness

20 pts — Response time and uptime

Capability Depth

15 pts — MCP tools, structured data

Stability

10 pts — Accountability metadata, attestation

Average readiness across the index: —/100

Check your service's readiness score

Trust & Security Trust tiers, signals, and safe defaults

Discovery is not endorsement. The index tells you what exists — your agent decides whether to trust it.

Access Classification

Every service is probed and classified — free, API key, OAuth, or x402. Know the requirements before calling.

Risk Scoring

Scored 0–100 based on TLS validity, response integrity, and community reports. High-risk services filtered by default.

Response Sanitization

Clean service responses before they reach your LLM. Strips HTML, scripts, and prompt injection patterns.

Community Reporting

Flag malicious services via POST /report. Reported services are marked in search results.

No Middleman

The index serves metadata only. Your agent calls services directly — no proxy, no data stored.

Verify Before Calling

Check TLS certificates, scan for injection patterns, and detect suspicious redirects before your agent sends data.

Trust Tiers:
Discovered — Crawler found this service and it responded.
Verified — Alive across 3+ crawl cycles, valid TLS, structured data.
Attested — Operator proved domain ownership via DNS TXT or well-known file.

How to attest your service

Attestation proves you own the domain. It raises your trust tier to Attested and boosts your readiness score by up to 10 points.

Step 1: Request a code

curl -X POST https://montexi.com/attest \
  -H "Content-Type: application/json" \
  -d '{"endpoint":"https://api.example.com"}'

Step 2: Prove ownership

Either add a DNS TXT record or serve the code at:

/.well-known/montexi-verify.txt

Step 3: Verify

curl -X POST https://montexi.com/attest/verify \
  -H "Content-Type: application/json" \
  -d '{"endpoint":"https://api.example.com","method":"well-known"}'

Declare accountability metadata

Services that declare their operator, terms of service, and support contact earn higher readiness scores. This metadata is surfaced in search results.

curl -X POST https://montexi.com/declare \
  -H "Content-Type: application/json" \
  -d '{
    "endpoint": "https://api.example.com",
    "operator": "Example Corp",
    "terms_url": "https://example.com/terms",
    "privacy_url": "https://example.com/privacy",
    "support_email": "support@example.com"
  }'

What the index does not do: Audit source code, verify operator identity, guarantee uptime, or store your data. Always verify before sending sensitive data to any discovered service.

SDK safe defaults: max_risk: 50, safeFetch() with 10s timeouts, sanitizeResponse() for LLM-bound content. Safety on by default.

How It Works Three steps from publication to execution

Services declare

Publish a .well-known/agent-card.json, .well-known/agent.json, or OpenAPI spec on your domain.

The index discovers

The multi-source crawler probes thousands of domains every 6 hours, checking multiple discovery paths per domain.

Agents query and execute

Install the SDK, search by capability, get live endpoints, interact directly.

Principles Design philosophy

Stateless

No sessions, no accounts, no stored state.

Non-authoritative

The index suggests, services define truth, the agent decides.

Payment-agnostic

Works with free services, x402, Stripe, or any payment model.

Multi-protocol

Indexes A2A, MCP, OpenAPI, and agent.json in a single search.

Accountable

Services declaring operator, terms, and support earn higher readiness scores.

Montexi