Live on Solana Devnet

The trust layer
for AI agents

Vouch is on-chain identity and behavioral reputation for AI agents on Solana. Free to use. Any protocol can verify agent trust via CPI. Any AI model can discover agents via MCP. Register once, be trusted everywhere.

Start Integrating How it Works
4
Open Standards
10
MCP Tools
2
CPI Instructions
6
Agent Capabilities

Built on proven, interoperable standards

Solana W3C DID v1.1 A2A Agent Card v0.3 MCP 2025-11-25 W3C VCs v2.0 Anchor USDC (SPL)

AI agents are transacting autonomously.
Trust hasn't caught up.

An AI agent is just a wallet. No identity, no track record, no accountability. When agents hire, buy from, or negotiate with other agents, they're operating on blind faith.

?

No Identity

Who is this agent? Who operates it? What are its capabilities? A wallet address tells you nothing.

!

No Trust Signal

Should I trust this agent with $10 or $10,000? There is no mechanism to distinguish a legitimate operator from a disposable bot.

0

No Track Record

An agent could have completed 1,000 successful transactions or failed on every one. There is no behavioral history to check.

X

No Accountability

If an agent misbehaves, it can abandon its wallet and spin up a fresh identity for free. Reputation laundering is trivial.

Three steps to a universally trusted agent

Free to use. No KYC. Fully on-chain. Once registered, your agent is verifiable by any Solana program via CPI and discoverable by any AI model via MCP — instantly.

1

Register Identity

Give your agent a name, a service category, and declare its capabilities. Link your A2A Agent Card for cross-platform discovery. In seconds, your agent has an on-chain identity, a W3C DID, and is resolvable by any protocol in the ecosystem.

On-chain PDA + W3C DID + A2A Agent Card
2

Stake USDC

Deposit USDC to signal commitment — your funds stay yours and are fully withdrawable. Trust tier upgrades automatically: $100 for Basic, $1K for Standard, $10K for Premium. A 7-day withdrawal cooldown ensures other protocols can rely on your stake. Vouch never takes your money.

Economic trust signal verifiable by any protocol
3

Build Reputation

As your agent transacts across protocols, each one reports how it behaves. Your reputation score (0–1000) follows you everywhere. Stake + reputation together form a dual trust signal that's expensive to fake and impossible to reset.

Portable cross-protocol reputation score
Result
Any Solana program can verify your agent via CPI — one instruction
Any AI model can discover and verify your agent via MCP — one tool call
Any identity system can resolve your agent via W3C DID — globally interoperable
Third parties can issue Verifiable Credentials attesting to audits, compliance, and capabilities

Built on open standards. Discoverable everywhere.

Vouch does not create another silo. Every registered agent is automatically discoverable through four established standards. We build on proven, interoperable infrastructure — not proprietary lock-in.

W3C Decentralized Identifiers

DID v1.1 · did:sol method

Every agent gets did:sol:<pubkey> — a globally unique, self-sovereign identifier. Full W3C DID v1.1 documents with Ed25519 verification, authentication, and service endpoints. Any system that speaks W3C DID can resolve Vouch agents without custom integration.

Ed25519 Verification Authentication Service Endpoints SSI Interop

A2A Agent Cards

Agent Card v0.3

Industry-standard agent discovery. Agents store a URL + SHA-256 hash on-chain. Vouch's indexer verifies card integrity every 10 minutes — built-in tamper detection. Links to DID Document service endpoints.

On-chain Hash 10-min Integrity Check Tamper Detection Trust Data Extension

Model Context Protocol (MCP)

MCP 2025-11-25 · SSE transport

Full MCP 2.0 server with SSE transport. Any MCP-compatible AI model can connect and autonomously discover, verify, and register agents. 10 tools — read and write — zero integration code needed.

lookup_agent verify_agent list_agents check_reputation resolve_did check_credentials prepare_register prepare_update prepare_deposit submit_tx

W3C Verifiable Credentials

VC v2.0

Third parties issue verifiable attestations about agents. Vouch indexes and structurally verifies them. Trust what has been independently verified, not just what is self-claimed.

Security Audit Identity Capability Compliance

Skin in the game. Visible to everyone.

Agents stake USDC as collateral behind their identity. Higher stake = stronger trust signal. Protocols decide what tiers they require. The economics make honesty the only rational strategy.

👁
Observer
$0
USDC staked
  • On-chain identity
  • Reputation tracking
  • DID + Agent Card
  • No economic commitment
💠
Basic
$100
USDC staked
  • Entry-level commitment
  • Stake lockable during disputes
  • 7-day withdrawal cooldown
  • Signals intent to transact
💎
Premium
$10,000
USDC staked
  • Maximum commitment
  • Stake lockable during disputes
  • 7-day withdrawal cooldown
  • Strongest credibility signal

Tiers are a credibility signal, not a feature gate. What protocols do with that signal is entirely up to them. Auto-tier upgrades as deposits increase. Admin can freeze stake during investigations.

Behavioral track record that follows the agent

Trust is not just about how much money an agent has staked. It is about how that agent has behaved. Vouch combines both signals into the strongest trust metric in the ecosystem.

720
/ 1000
Starting: 500
Max: 1,000
Min: 0
Weighted scoring

Whitelisted Reporters CPI

Only whitelisted protocols can submit reputation signals. This happens via CPI (Cross-Program Invocation) — the reporting protocol calls Vouch's on-chain instruction directly. No off-chain oracles, no trust assumptions.

Weighted by Importance 1–10

Each reporter protocol is assigned an importance weight (1–10). A high-importance marketplace reporting fraud has more impact than a low-importance monitoring service reporting a ping failure.

Rate-Limited

Reputation reports are rate-limited per reporter per agent to prevent gaming. A reporter cannot spam positive signals to inflate an agent's score, or negative signals to tank a competitor.

Stake Freeze for Investigations

Admin can freeze an agent's stake during active investigations. The agent cannot withdraw while under review. Combined with deactivation records, this creates a comprehensive accountability system.

The Dual-Signal Trust Model

Neither stake nor reputation alone is sufficient. Together, they are far stronger than either signal alone.

USDC Stake
Proven commitment
+
Reputation
Behavioral track record
=
Vouch Trust
Verifiable confidence

Any Solana program can plug into Vouch

This is the core value proposition. Your protocol does not need to build its own trust system. Call verify_agent or check_agent_eligibility via CPI, and gate your logic on Vouch trust data. Two instructions. That's it.

1

verify_agent

Pass an agent's PDA. Returns whether the agent is active, its tier, stake amount, and reputation score. Use this to gate access to your protocol's instructions.

2

check_agent_eligibility

Pass an agent PDA plus your requirements (minimum tier, minimum reputation, required capabilities). Returns a boolean. One instruction, one answer: eligible or not.

+

Report Reputation (Whitelisted)

If your protocol is whitelisted as a reporter, you can submit behavioral signals about agents via CPI. Positive or negative, weighted by your assigned importance level.

0

Zero Infrastructure Cost

No databases, no APIs, no oracles. All trust data lives on-chain. You pass the account, Vouch validates it. Composability the way Solana was designed to work.

Rust (Anchor CPI)
vouch-cpi
use vouch_cpi::{AgentIdentity, AgentTier};

// Gate your instruction on Vouch trust
require_vouch_active!(ctx.accounts.agent);
require_vouch_tier!(
    ctx.accounts.agent,
    AgentTier::Standard
);

// Check specific eligibility
let eligible = check_agent_eligibility(
    &ctx.accounts.agent,
    AgentTier::Basic,    // min tier
    600,                   // min reputation
    CAP_SELL | CAP_NEGOTIATE // required caps
);

// Unverified agents are rejected
// automatically. No external calls.

Integrate in minutes. Two paths.

On-chain: Use the Rust CPI crate to gate your Solana program on Vouch trust. Off-chain: Use the TypeScript SDK to verify agents or the MCP server for AI-native access.

Rust (Anchor CPI)
vouch-cpi
use vouch_cpi::{AgentIdentity, AgentTier};

// In your instruction handler:
require_vouch_active!(ctx.accounts.agent);
require_vouch_tier!(
    ctx.accounts.agent,
    AgentTier::Standard
);

// Unverified agents are rejected
// automatically. No external calls.
TypeScript
@vouch/verify
import { fetchAgent, isActive,
  hasTier, meetsRequirements
} from "@vouch/verify";

const agent = await fetchAgent(connection, pk);

if (isActive(agent) && hasTier(agent, "Standard")) {
  // Agent is verified. Proceed.
}
DID Resolution
did:sol
// Every Vouch agent has a W3C DID
const did = deriveAgentDid(ownerPubkey);
// "did:sol:7xKXtg2CW87d97TXJSDp..."

// Resolve DID to agent identity
const agent = await registry.resolveAgentByDid(did);

// Generate W3C DID Document
const doc = await registry.generateDidDocument(owner);
// Returns full v1.1 DID Document with
// verificationMethod, authentication,
// and service endpoints
MCP Tool Call
MCP 2.0
// Any AI model can register via MCP
const prep = await mcp.callTool(
  "prepare_register_agent", {
    ownerPubkey: wallet.toBase58(),
    name: "MyAgent",
    serviceCategory: "commerce",
    agentCardUrl: "https://...agent.json",
    agentCardHash: [...sha256]
  }
);
// Sign client-side, then submit
tx.sign([wallet]);
await mcp.callTool("submit_transaction", {
  txId: prep.txId, signedTx: tx
});

Cryptographic on-chain identity for every agent

Each agent gets a unique PDA derived from its owner wallet. Identity is permanent, verifiable, and linked to every open standard in the ecosystem.

ON-CHAIN

Unique PDA Identity

Every agent gets a Program Derived Address from its owner wallet. This is the agent's permanent on-chain identity — verifiable by any program, readable by any client.

Buy Sell Negotiate Release Escrow Delegate Arbitrate
DID

W3C Decentralized Identifier

Every agent automatically gets did:sol:<pubkey>. Full W3C DID v1.1 compliant documents with Ed25519 verification methods, authentication, and service endpoints. Interoperable with the entire SSI ecosystem.

A2A

Agent Card (URL + Hash)

Agents store an A2A Agent Card URL and its SHA-256 hash on-chain. Vouch's indexer verifies card integrity every 10 minutes — tamper detection built in. Follows the A2A Agent Card v0.3 spec.

ANTI-SYBIL

Deactivation Records

When an agent is deactivated, a permanent record is stored on-chain. This prevents reputation laundering — bad actors cannot delete their history and start fresh with a clean identity.

Built where agents actually transact

Agent-to-agent commerce requires sub-second finality, negligible fees, and deep stablecoin liquidity. Solana is the only chain where all three converge.

400ms

Finality

Agents transact at machine speed. No waiting for block confirmations.

$0.001

Per Transaction

Micro-transactions are economically viable. Register, stake, and verify for fractions of a cent.

SPL USDC

Stablecoin

Deep liquidity. No token risk. The standard for on-chain dollars. All Vouch staking is denominated in USDC.

The agent economy needs a trust layer. Build on Vouch.

Whether you are building an AI agent that needs identity, or a protocol that needs to verify agents — Vouch gives you the infrastructure. CPI on-chain. MCP off-chain. Open standards everywhere.