Ventru.net docs

The simple version first. The protocol details after.

Ventru.net helps AI agents find real businesses, verify who is safe, get a signed action plan, pay for API usage, and report whether the job worked.

The simple version

The old web: people click websites, compare tabs, fill out forms, and hope the business is real.

The Ventru network: an AI agent asks for the safest route to complete a task. Ventru returns a signed packet with who to contact, why they ranked, what the agent must ask, how payment or fees work, and how to report the outcome.

Think of Ventru like search, verification, payment, and reputation infrastructure for AI execution. Search engines ranked pages. Ventru ranks executable outcomes.

Why this needs to exist

The web was designed for people, not agents. Agents waste time scraping pages, miss stale details, and cannot reliably tell which provider is real. Businesses also need a clean way to publish: “Here is my agent profile, here are my actions, here is my verification, here is how an agent can pay or book.”

For agents

Ventru gives machines structured discovery, trust proof, action endpoints, payment terms, and receipt templates.

For businesses

Ventru gives companies a way to become visible, trusted, and callable by AI agents before the agent web becomes mainstream.

The Ventru loop

Discovery → TrustGraph → AgentRank → Action Capsule → x402 Rail → Execution → Outcome API → Telemetry

1. Discovery

WebProbe checks a normal site for agent native signals like llms.txt, agent.json, OpenAPI, MCP, x402, schema data, and action endpoints.

2. TrustGraph

Ventru tracks proof: domain ownership, license, insurance, freshness, ratings, response SLA, and verification status.

3. AgentRank

Providers are scored by relevance, trust, actionability, freshness, and budget fit.

4. Action Capsule

The winning route is packaged into a signed, expiring payload that an agent can execute safely.

5. Outcome API

After the action, the agent reports accepted, completed, failed, disputed, response time, and satisfaction signals. That telemetry improves future ranking.

Action Capsules

An Action Capsule is the core proprietary primitive. It is not a blog post, not a directory card, and not a static listing. It is a short lived action packet for autonomous agents.

Contains

User intent, ranked providers, trust policy, commercial disclosure, executable route bodies, required fields, receipt template, expiry, and HMAC signature.

Why it matters

Agents need safe execution paths, not another page to scrape. Capsules turn search into action.

POST /api/capsule
{
  "intent": "emergency roof repair palm beach",
  "budget": 1000,
  "urgency": "today"
}

TrustGraph and AgentRank

TrustGraph is the proof layer. AgentRank is the scoring layer. Together they tell agents which route is safest and most actionable.

x402 payment rail

Agents should be able to discover a price, pay automatically, and receive a paid API result without a sales call. Ventru exposes machine readable payment terms through /.well-known/x402.json and paid resources like /api/paid/capsule.

GET /.well-known/x402.json
POST /api/paid/capsule
Header: X-PAYMENT: demo-receipt

The current MVP accepts demo payment headers. Production should verify settled x402 facilitator receipts before serving paid resources.

Endpoint map

Discovery

POST /api/discover
GET /.well-known/agent.json
GET /.well-known/agent-index.json
GET /llms.txt

Routing

GET /api/search
POST /api/intent
POST /api/capsule
POST /api/paid/capsule

Trust

GET /api/verify/{slug}/challenge
GET /api/verification-plans
POST /api/outcome

Agent interoperability

GET /openapi.json
GET /.well-known/mcp.json
GET /.well-known/ai-plugin.json
GET /.well-known/ventru-protocol.json

How Ventru makes money