# AgentScore

> AgentScore is commerce infrastructure for selling to AI agents. An AI agent is a faceless buyer, so the hard part is knowing who's behind it: AgentScore verifies the real human behind every agent (KYC, age, sanctions, jurisdiction) and returns a trust decision, then takes the payment and keeps the merchant compliant. Four primitives ship today: AgentScore Pay (universal agent CLI), AgentScore Commerce (merchant SDK across 11 frameworks), AgentScore Passport (free buyer-side KYC credential), and AgentScore Gate (compliance middleware bundled in Commerce). AgentScore Forward-Deployed Engineering is available for merchants that want AgentScore to build the integration end to end.

## What ships today

- **AgentScore Pay**: universal CLI. One shell command pays any x402 or MPP merchant. Native x402 (Base), MPP (Tempo, Solana), Stripe shared payment token handoff. Encrypted local keystore, MCP server. Free for buyers and agents.
- **AgentScore Commerce**: merchant SDK across 11 frameworks (Express, Hono, Fastify, Next.js, Web Fetch on Node; FastAPI, Flask, Django, AIOHTTP, Sanic, ASGI on Python). Multi-rail x402/MPP builder, Stripe multichain glue, x402 v1/v2 version handling, discovery manifests at `/skill.md` and `/.well-known/mpp.json`.
- **AgentScore Passport**: buyer-side identity. Verify KYC, age, sanctions, and jurisdiction once via Stripe Identity; the credential follows the buyer across every gated merchant. Free for buyers, forever. 24-hour access token + 90-day refresh, rotated silently.
- **AgentScore Gate**: compliance middleware bundled in Commerce. KYC, age (18+/21+), OpenSanctions screening, jurisdiction. Auto-mints verification sessions on missing identity so agents bootstrap inline. For regulated verticals (wine, age-restricted goods, export-controlled software, PII).
- Plus: **AgentScore Forward-Deployed Engineering**, for merchants without engineers to spare. AgentScore engineers embed and ship the integration end-to-end. Martin Estate (agents.martinestate.com, wine) and Sayer & Stone (agents.sayerandstone.com, lab-grown diamond jewelry) are the canonical examples.

## Audience

- **Buyers running agents**: use AgentScore Pay to pay merchants and AgentScore Passport to verify identity once.
- **Merchants**: drop AgentScore Commerce into your stack to accept multi-rail agent payments and gate compliance per route. Or work with our Forward-Deployed Engineering team.

## Pages

- [Home](https://www.agentscore.com): commerce infrastructure for AI agents
- [Identity](https://www.agentscore.com/identity): The identity layer for agentic commerce. Resolve the verified operator behind every agent; private by design
- [For Merchants](https://www.agentscore.com/merchants): Sell to AI agents. Self-serve SDK or Forward-Deployed Engineering
- [For Buyers](https://www.agentscore.com/buyers): Verify once, transact at every gated merchant
- [Pricing](https://www.agentscore.com/pricing): Free for agents, metered for merchants, FDE for high-touch integrations
- [Case Studies](https://www.agentscore.com/case-studies): Live storefronts in production today
- [Martin Estate Case Study](https://www.agentscore.com/case-studies/martin-estate): Wine, regulated commerce, full FDE narrative
- [Sayer & Stone Case Study](https://www.agentscore.com/case-studies/sayer-and-stone): Lab-grown diamond jewelry, selling to AI agents
- [About](https://www.agentscore.com/about): Company, partners, and approach
- [Blog](https://www.agentscore.com/blog): Notes on agentic commerce, identity, payments, and compliance from the AgentScore team
- [Careers](https://www.agentscore.com/careers): How we work and how to apply; based in San Francisco
- [Contact](https://www.agentscore.com/contact): Forward-Deployed Engineering, enterprise plans, partnerships
- [Get a Passport](https://www.agentscore.com/dashboard/verify): Identity verification flow (KYC)
- [API Docs](https://docs.agentscore.com): Full API reference and integration guides
- [Blog feed](https://www.agentscore.com/feed.xml): RSS for the blog

## Marketplace (the agentic economy, measured)

Public data surfaces. Every figure is settled on-chain activity or an independently gathered signal, each carrying the date it was observed; missing data renders as absent rather than as zero.

- [Marketplace](https://www.agentscore.com/marketplace): Settled agent payment volume and count over time, sliceable by chain, protocol, and payee registry
- [Agent directory](https://www.agentscore.com/marketplace/directory): Every cataloged agent and service, one row each, with settled payments, compliance claims, endpoint health, identity, and domain posture
- [Transaction explorer](https://www.agentscore.com/marketplace/transactions): Individual settled agent payments, filterable by chain, facilitator, protocol, payee registry, and address
- [Economy methodology](https://www.agentscore.com/marketplace/data): What "settled" means, the on-chain sources, and per-source freshness

## Specification

Machine-readable declarations a gated merchant publishes, and the AgentScore bindings for each payment rail in UCP.

- [Identity capability](https://www.agentscore.com/specification/identity): `com.agentscore.identity`, the UCP profile capability declaring an AgentScore-gated merchant
- [x402 payment handler](https://www.agentscore.com/specification/payment-handlers/x402): EVM scheme on Base today; Solana and Stellar slot in as additional networks
- [MPP payment handler](https://www.agentscore.com/specification/payment-handlers/mpp): Tempo USDC and Solana USDC today, more MPP networks by config
- [Stripe SPT payment handler](https://www.agentscore.com/specification/payment-handlers/stripe_spt): Agent-minted, merchant-scoped shared payment tokens
- [Identity policy schema](https://www.agentscore.com/schemas/ucp/com-agentscore-identity-v1.json): The policy JSON Schema itself

## Hivemind

- [Hivemind](https://hivemind.agentscore.com): AI agents built into your company as coworkers with real jobs, working from one shared brain, with a human yes before anything ships
- [Meet the Bees](https://hivemind.agentscore.com/meet-the-bees): The agents that work inside AgentScore, each with a real job

## Reputation signal (optional)

Beyond the identity decision, `GET /v1/reputation` returns an optional on-chain trust score (0 to 100) and letter grade (A to F) for a wallet. It is a supplementary signal, not the gate; the identity layer and the commerce stack are the lead.

## API

- Base URL: https://api.agentscore.com
- Docs: https://docs.agentscore.com
- Authentication: `X-API-Key` header, key obtained from the dashboard (free tier available)

## Endpoints

- `GET /v1/reputation/{address}`:Reputation lookup for any wallet address
- `POST /v1/assess`:Trust assessment with policy evaluation and reason codes
- `POST /v1/sessions`:Create verification session for identity bootstrapping
- `GET /v1/sessions/{token}`:Poll session status; returns one-time `operator_token` on verified
- `POST /v1/credentials`:Create operator credential
- `GET /v1/credentials`:List active credentials
- `DELETE /v1/credentials/{id}`:Revoke credential
- `POST /v1/credentials/wallets`:Report a signer wallet captured at payment time

## Pricing

See [agentscore.com/pricing](https://www.agentscore.com/pricing) for current tier names, quotas, and rate limits. Buyers and agents always free. Merchants pay for API usage. Forward-Deployed Engineering is engagement-based.

## Integrations

- [TypeScript SDK](https://docs.agentscore.com/integrations/typescript): `npm install @agent-score/sdk`. Typed client for Node and edge runtimes.
- [Python SDK](https://docs.agentscore.com/integrations/python): `pip install agentscore-py`. Sync + async client.
- [Node Commerce SDK](https://docs.agentscore.com/integrations/node-commerce): `npm install @agent-score/commerce`. Identity middleware (Hono, Express, Fastify, Next.js, Web Fetch), payment helpers, 402 builders, discovery, Stripe multichain.
- [Python Commerce SDK](https://docs.agentscore.com/integrations/python-commerce): `pip install agentscore-commerce`. Identity middleware (FastAPI, Flask, Django, AIOHTTP, Sanic, ASGI), payment helpers, 402 builders, discovery, Stripe multichain.
- [Pay CLI](https://docs.agentscore.com/integrations/pay-cli): `npm install -g @agent-score/pay`. Agent-side wallet, payment, identity. Doubles as MCP server via `agentscore-pay --mcp`.

## Agent-side CLI

`@agent-score/pay` is the agent-facing CLI. Wallet management, x402/MPP payments, identity (assess, sessions, credentials, associate-wallet, reputation), all in one binary. Doubles as a stdio MCP server (`agentscore-pay --mcp`) and exposes an LLM manifest (`agentscore-pay --llms`). Agents in Claude, GPT-4o, Gemini, or any MCP-compatible runtime can register the CLI with `agentscore-pay mcp add` and call every command as a tool.

## Partners

AgentScore orchestrates the existing ecosystem rather than competing with it. Stripe for identity (Stripe Identity) and card payments. Coinbase, Tempo, and Solana for crypto rails. x402 (x402.org) and MPP (paymentauth.org) for protocol-level interoperability. OpenSanctions for sanctions screening.

## In production today

[Martin Estate](https://agents.martinestate.com) sells wine to AI agents. Bottles ship to verified buyers with per-state tax calculated at checkout. The API is gated by a single `@agent-score/commerce` middleware call that enforces KYC, age 21+, sanctions, and US-only jurisdiction. Payment via x402 USDC on Base, MPP USDC on Tempo and Solana, or Stripe shared payment token. Built by AgentScore Forward-Deployed Engineering.

[Sayer & Stone](https://agents.sayerandstone.com) sells lab-grown diamond fine jewelry (bracelets and earrings, made to order) to AI agents. No identity gate on `/purchase` (jewelry is unregulated and full KYC is disproportionate to the goods); every payment rail authenticates via its own native credential. Payment via x402 USDC on Base, MPP USDC on Tempo, or Stripe shared payment token. Built by AgentScore Forward-Deployed Engineering.

## Optional

- [Terms of Service](https://www.agentscore.com/terms)
- [Privacy Policy](https://www.agentscore.com/privacy)
- [Data Use Policy](https://www.agentscore.com/data-use-policy)
- [GitHub](https://github.com/agentscore)
- [X/Twitter](https://x.com/agentscoretrust)