Skip to main content
Sly speaks six protocols that let AI agents transact with merchants, services, and each other. You integrate once, and the platform adapts to whichever protocol a counterparty uses.
ProtocolOriginBest forTypical flow
UCPSly-nativeMerchant checkout, tokenized commerce, settlement rulesMerchant issues UCP token → agent exchanges for settlement
ACPStripe + OpenAIAgentic checkout sessions with card networksAgent opens checkout → completes payment → merchant confirms
AP2GooglePre-authorized spending mandatesAgent presents mandate → merchant executes within policy
x402CoinbaseMicropayments for APIs (HTTP 402 revival)API returns 402 → agent pays → retries with proof
A2AGoogleAgent-to-agent task delegation and discoveryAgent A sends task to Agent B → gets deliverable → pays
MCPAnthropicTool catalogs and resources for LLM agentsLLM discovers tools → invokes → receives results

​
How to pick

Most integrations use one or two protocols aligned to their partner ecosystem. Start here:
  • Selling on Stripe / OpenAI marketplaces → ACP
  • Selling on Google Cloud / Gemini ecosystem → AP2 and A2A
  • Monetizing APIs via micropayments → x402
  • Building a traditional checkout with agentic capabilities → UCP
  • Shipping Claude Desktop / agent tools → MCP
If in doubt, see the decision tree.

​
One SDK, all protocols

The @sly_ai/sdk exposes each protocol behind a dedicated client: 
import { Sly } from '@sly_ai/sdk';
const sly = new Sly({ apiKey: process.env.SLY_API_KEY });

// Pick your protocol
await sly.ucp.createToken({...});
await sly.acp.createCheckout({...});
await sly.ap2.createMandate({...});
await sly.x402.pay({...});
await sly.a2a.sendTask({...});
await sly.mcp.listTools();

​
Composition

Protocols compose — you can issue a UCP token that settles on-chain via x402, or accept an AP2 mandate that executes through an ACP checkout. See /v1/composition and the composition guide for cross-protocol routing.

​
Discovery

Every protocol Sly supports is self-describing: 
curl https://api.getsly.ai/v1/protocols
Returns the full list of enabled protocols for your tenant, their versions, and machine-readable capability specs. Agents can use this to negotiate at runtime.