The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the AgentServices listing page.
The paid API layer for AI agents — data, intelligence, inference, and media via x402 micropayments
Canonical project: https://github.com/vbkotecha/agentservices-api Official site: https://agentservices.to License: Apache-2.0
AgentServices is an independent project. It is not affiliated with, endorsed by, or a replacement for any other x402 market-data provider.
Live at: agentservices.to | MCP Server: https://agentservices.to/mcp (Streamable HTTP) | Discovery: /.well-known/x402
AgentServices is the monetized API layer for AI agents. No API keys, no subscriptions — agents pay per-request with USDC on Base using the x402 payment protocol.
60+ endpoints across crypto data, market intelligence, DeFi analytics, on-chain analytics, 400+ LLM models, image generation, text-to-speech, portfolio intelligence, and dispute resolution. 12 are free. 48+ are paid via x402 (from $0.002 to $0.25 per call).
/v1/chat/completions — GPT, Claude, Gemini, DeepSeek, Grok, Llama, and moremodel: "auto" and the gateway classifies your task and picks the cheapest model that handles it/v1/images/generations via gpt-image-2 ($0.05)/v1/audio/speech with natural voices ($0.05)chat, generate_image, text_to_speechUse the buyer quickstart for the complete path. The exact next action is to run the no-credential discovery check:
Then follow the same buyer journey:
The four proofs have different limits: discovery and free SDK verify no-spend access; the paid SDK proof verifies challenge shape only; the receipt builder hashes buyer-held evidence and does not verify settlement. None of them claims adoption, settlement, or revenue.
For activation measurement definitions, see Activation Metrics. To probe the live funnel, run python3 examples/check_activation_funnel.py.
Use the buyer quickstart for the complete path. The exact next action is to run the no-credential discovery check:
Then follow the same buyer journey:
The four proofs have different limits: discovery and free SDK verify no-spend access; the paid SDK proof verifies challenge shape only; the receipt builder hashes buyer-held evidence and does not verify settlement. None of them claims adoption, settlement, or revenue.
| Endpoint | Description |
|---|---|
GET /v1/prices?symbols=BTC,ETH | Current crypto prices (CoinGecko) |
GET /v1/trending | Trending tokens by market activity |
GET /v1/global | Global market cap, volume, BTC dominance |
GET /v1/fear-greed | Crypto Fear & Greed sentiment index |
GET /v1/gas | Current gas prices on Base/Ethereum |
GET /v1/predictions | AI-generated market predictions |
GET /v1/news | Latest crypto news headlines |
GET /v1/social/trending | Trending social sentiment topics |
GET /v1/geo?ip=1.2.3.4 | IP geolocation lookup |
GET /v1/swap/quote?from=&to=&amount= | DEX swap quote (0x integration) |
GET /v1/policies | List dispute resolution policy templates |
GET /v1/trade/hyperliquid/bootstrap | Hyperliquid agent-sign bootstrap (approveAgent flow) |
POST /v1/trade/hyperliquid/order | Forward agent-signed HL order (policy-gated) |
POST /v1/trade/hyperliquid/cancel | Forward agent-signed HL cancel |
GET /v1/trade/hyperliquid/order | HL order status (query: user, oid) |
GET /v1/trade/hyperliquid/order/{id} | HL order status (path param) |
GET/PUT /v1/trade/hyperliquid/policy | Execution leash (max notional, coin allowlist, kill switch) |
POST /v1/trade/hyperliquid/paper/order | Paper/sim order for agent training |
POST /v1/trade/hyperliquid/eval/order | Policy pass/fail eval (training gym) |
GET /health | API health check |
AgentServices is building venue trade doors under /v1/trade/{venue}/…. Hyperliquid ships first; more venues and a venue-neutral /v1/trade/execute router follow.
Each door is a policy leash + forward — not a smarter router. We do not claim better fills than the venue direct. Execution is free at the call (no x402 on the order path). Builder fees are omitted so routing through us is not more expensive than going direct.
We never collect venue API keys. Agents sign orders locally with an HL-approved agent wallet (approveAgent on the main wallet is the one human bootstrap step). AgentServices policy-checks (max notional, coin allowlist, kill switch) and forwards the signed payload.
Request bodies accept market_type: spot, perp, or future. Hyperliquid implements perp and spot today; unsupported types return a machine-readable market_type_not_supported error.
| MCP tool | HTTP equivalent |
|---|---|
trade_hyperliquid_order | POST /v1/trade/hyperliquid/order |
trade_hyperliquid_cancel | POST /v1/trade/hyperliquid/cancel |
trade_hyperliquid_order_status | GET /v1/trade/hyperliquid/order |
trade_hyperliquid_get_policy / trade_hyperliquid_set_policy | GET/PUT /v1/trade/hyperliquid/policy |
trade_hyperliquid_paper_order | POST /v1/trade/hyperliquid/paper/order |
trade_hyperliquid_eval_order | POST /v1/trade/hyperliquid/eval/order |
Legacy hl_* MCP tool names remain as aliases. See GET /v1/trade/hyperliquid/bootstrap for the approveAgent signing model.
Roadmap: /v1/trade/{venue} for additional venues; /v1/trade/execute for venue-neutral routing later.
| Endpoint | Price | Description |
|---|---|---|
GET /v1/indicators/BTC | $0.02 | Technical indicators (RSI, MACD, Bollinger Bands, ATR, Support/Resistance) |
GET /v1/yields | $0.02 | Top DeFi yield pools by TVL |
GET /v1/metadata?url= | $0.01 | URL metadata extraction and unfurling |
GET /v1/search?q= | $0.01 | Web search for crypto/market information |
GET /v1/onchain/:address | $0.02 | On-chain analytics for any address |
GET /v1/onchain/:address/tokens | $0.03 | Token holdings for any address |
| Endpoint | Price | Description |
|---|---|---|
GET /v1/token-risk/{token} | $0.03 | Snapshot-based volatility, liquidity-proxy, and market-cap risk score |
GET /v1/crypto-signals | $0.04 | Aggregated buy/sell signals across multiple indicators |
GET /v1/yield-comparison | $0.03 | Compare yields across protocols with risk-adjusted returns |
GET /v1/hn-sentiment | $0.02 | Hacker News sentiment analysis for tech topics |
GET /v1/npm-stats/:package | $0.02 | NPM package download stats and trends |
GET /v1/github-trending | $0.02 | Trending GitHub repos by language/topic |
GET /v1/marketing-intel | $0.05 | Marketing intelligence: competitors, content gaps, ad copy |
| Endpoint | Price | Description |
|---|---|---|
GET /v1/research?q= | $0.05 | Deep research: search + extract + synthesize in one call |
GET /v1/portfolio?symbol=BTC | $0.10 | Portfolio intelligence: price + signal + risk + sentiment + verdict |
GET /v1/defi-strategy | $0.25 | DeFi strategy report: top yields + TVL + comparison + risk flags |
GET /v1/market-pulse | $0.05 | Market pulse: fear-greed + trending + news + social + whales + global |
GET /v1/onchain-overview | $0.15 | On-chain overview: whales + exchange flows + stablecoin flows + correlation + DeFi TVL |
| Endpoint | Price | Description |
|---|---|---|
POST /v1/inference | $0.03 | LLM inference (gpt-5.4/5.4-mini/5.5) — chat completions |
POST /v1/complete | $0.03 | Text completion (CodexSale proxy) |
| Endpoint | Price | Description |
|---|---|---|
POST /v1/disputes | $0.05 | AI-powered dispute resolution (7 policy templates) |
13+ MCP tools available: crypto_prices, trending_tokens, global_market, gas_prices, market_predictions, crypto_news, social_trending, technical_indicators, defi_yields, search_web, token_risk, crypto_signals, onchain_analytics, deep_research, portfolio_intelligence, defi_strategy, market_pulse, onchain_overview
AgentServices includes an AI-powered dispute resolution system with 7 policy templates:
| Policy | Use Case |
|---|---|
freelance-delivery | Freelancer vs client delivery disputes |
milestone-payment | Milestone-based project payment disputes |
sla-monitoring | Service level agreement violations |
api-quality | API response quality / uptime disputes |
bug-bounty | Bug bounty validity disputes |
scope-dispute | Project scope creep disputes |
physical-commerce | Physical goods transaction disputes |
Wallet agents continue to pay via x402 on REST — unchanged. Humans connecting through ChatGPT Developer Mode or Claude custom connectors can use Google OAuth + Stripe prepaid credits on MCP.
| Rail | Who | How |
|---|---|---|
| x402 | Wallet agents | Unauthenticated REST → HTTP 402 → USDC on Base |
| Credits | Logged-in humans | Google OAuth on MCP → deduct same USD price from prepaid balance |
MCP URL for ChatGPT: https://agentservices.to/mcp
Copy .env.example. Required to enable the human door:
| Variable | Purpose |
|---|---|
GOOGLE_CLIENT_ID | Google OAuth client ID |
GOOGLE_CLIENT_SECRET | Google OAuth client secret |
OAUTH_JWT_SECRET | Signs MCP bearer tokens (or use SESSION_SECRET) |
STRIPE_SECRET_KEY | Stripe API secret key |
STRIPE_WEBHOOK_SECRET | Verifies checkout.session.completed webhooks |
STRIPE_PRICE_CREDITS_10 | Optional Stripe Price ID for $10 pack (otherwise hardcoded) |
PUBLIC_BASE_URL | Canonical host, e.g. https://agentservices.to |
Prepaid credit balances are stored in Stripe Customer Balance (one Stripe Customer per Google sub). No Redis, Postgres, or other database is required.
Without Google/Stripe vars the API boots in x402-only mode.
Add both hosts:
https://agentservices.to/oauth/google/callbackhttps://api.agentservices.to/oauth/google/callbackPoint Stripe to:
https://agentservices.to/billing/webhookhttps://api.agentservices.to/billing/webhookEvent: checkout.session.completed
to.agentservices/agentservicesagentservices.to + api.agentservices.to, auto-deploy from main)AgentServices includes a built-in Coinbase AgentKit action provider (agentkit/ directory). This lets any AgentKit agent use AgentServices APIs with typed, documented actions:
18 actions covering free (prices, fear-greed, trending, gas) and paid endpoints (indicators, DeFi yields, portfolio intelligence, research, onchain analytics, AI inference, and more). See agentkit/README.md for full documentation.
MIT — Build on it, fork it, integrate it.