agents.txt

A proposed open standard for AI agent discovery and interaction.

robots.txt tells crawlers what they can access. agents.txt tells autonomous AI agents how they can participate.

As AI agents evolve from passive crawlers into active participants — bidding on auctions, making payments, consuming APIs — websites need a standard way to advertise what agent interactions they support.

Place an agents.txt file at your site root and optionally serve a JSON version at /.well-known/agents.

Full documentation & live validator: global-chat.io/agents-txt

---

Quick Start

Create a file at https://yoursite.com/agents.txt:

# agents.txt — Your Platform
# https://example.com/agents.txt

## IDENTITY
Name: Example Platform
URL: https://example.com
Description: A marketplace where AI agents can browse and purchase data feeds.
Contact: hello@example.com

## AGENT CAPABILITIES REQUIRED
Capability: HTTP requests (GET, POST)
Capability: JSON parsing
Capability: Crypto wallet (USDC on Base)

## AUTHENTICATION
Register: POST https://example.com/api/agents/register
Auth-Method: Bearer token

## ENDPOINTS
GET  /api/feeds/catalog    # Browse available data feeds
POST /api/feeds/purchase   # Purchase a feed subscription
GET  /api/feeds/:id/data   # Consume a purchased feed

## PAYMENT METHODS
Payment: USDC on Base (Chain ID: 8453)
Payment-Wallet: 0x1234...abcd
Min-Bid: 1.00 USDC

## MACHINE-READABLE
JSON: https://example.com/.well-known/agents

---

Spec Reference

Required Fields

Field	Description
Name	The name of the platform
URL	Canonical URL of the site
Description	What the platform offers to agents
Capability	A capability the agent must have (one per line, at least one required)
Register	The registration endpoint (METHOD https://...)
Auth-Method	How agents authenticate after registration

Optional Fields

Field	Description
Contact	Email or URL for human contact
Payment	Accepted payment method (currency + network)
Payment-Wallet	Wallet address for payments
Min-Bid	Minimum bid/payment amount
Rate-Limit	Request rate limit for agents
JSON	URL to the machine-readable JSON version

Sections

Organize your file with section headers for readability:

## IDENTITY
## AGENT CAPABILITIES REQUIRED
## AUTHENTICATION
## ENDPOINTS
## PAYMENT METHODS
## RATE LIMITS
## RULES
## MACHINE-READABLE

Endpoints

List API endpoints using standard HTTP method notation:

GET  /api/feeds/catalog     # Description
POST /api/agents/register   # Description

Comments

Lines starting with # are comments and are ignored by parsers.

---

JSON Format (/.well-known/agents)

For programmatic consumption, serve the same information as JSON:

{
  "spec_version": "1.0",
  "name": "Example Platform",
  "url": "https://example.com",
  "description": "A marketplace where AI agents can browse and purchase data feeds.",
  "contact": "hello@example.com",
  "capabilities_required": [
    "HTTP requests (GET, POST)",
    "JSON parsing",
    "Crypto wallet (USDC on Base)"
  ],
  "authentication": {
    "register": {
      "method": "POST",
      "endpoint": "https://example.com/api/agents/register",
      "body": {
        "agent_name": "string",
        "wallet_address": "string",
        "capabilities": ["string"]
      }
    },
    "method": "Bearer token"
  },
  "endpoints": [
    { "method": "GET", "path": "/api/feeds/catalog", "description": "Browse available data feeds" },
    { "method": "POST", "path": "/api/feeds/purchase", "description": "Purchase a feed subscription" }
  ],
  "payment": {
    "methods": [
      { "currency": "USDC", "network": "Base", "chain_id": 8453 }
    ],
    "wallet": "0x1234...abcd",
    "min_bid": "1.00 USDC"
  },
  "rate_limits": {
    "authenticated": "120/minute",
    "unauthenticated": "10/minute"
  }
}

---

How It Works

1. Agent discovers your site — fetches /agents.txt or /.well-known/agents
2. Agent registers — calls the registration endpoint with its name, wallet, and capabilities
3. Agent interacts — consumes feeds, places bids, or uses any listed endpoints
4. Payments settle on-chain — transactions are verified automatically

---

Validator

Validate your agents.txt file using the online validator:

global-chat.io/validate

Or use the TypeScript validator included in this repo:

import { validate } from './validator/agents-txt-validator';

const result = validate(myAgentsTxtContent);
console.log(result.valid);    // true/false
console.log(result.issues);   // array of errors, warnings, infos

---

Live Examples

• Global Chat — global-chat.io/agents.txt | JSON

Using agents.txt? Open a PR to add your site here.

---

Why agents.txt?

	robots.txt	agents.txt
Purpose	Controls crawler access	Describes agent interactions
Audience	Search engine bots	Autonomous AI agents
Actions	Allow/disallow crawling	Register, authenticate, transact
Payments	N/A	On-chain crypto payments
Authentication	N/A	API keys, tokens

---

Contributing

This is an open standard. Contributions are welcome:

• Spec improvements — open an issue or PR
• Validator enhancements — improve the TypeScript validator
• Examples — add your agents.txt implementation
• Integrations — build parsers in other languages

---

License

MIT