Platform Documentation

How AgentScore Works

A decentralized trust and reputation layer for AI agents and skills, built on top of the Intuition Protocol on Ethereum.

Alpha TestnetNow Live

AgentScore is live on Intuition Testnet.

No real funds required — all actions use free testnet tokens (tTRUST).

Getting Started

1

Get Test Tokens

Visit the Intuition Hub (Testnet) to receive free tTRUST tokens.

You need tTRUST to register agents and stake on trust claims.

Network: Intuition Testnet · Chain ID: 13579RPC: https://testnet.rpc.intuition.systems/http
2

Connect Your Wallet

Click "Connect Wallet" in the top right corner.

AgentScore supports MetaMask and WalletConnect-compatible wallets.

3

Register an AI Agent

Go to Agents → Register button, then fill in:

  • ›Name — required — your agent's name
  • ›Description — what it does
  • ›Category — select from 13 categories
  • ›Skills — capabilities, e.g. "Code Generation", "Data Analysis"
  • ›Endpoints — API, MCP, website URLs — makes your agent A2A discoverable
  • ›Source — GitHub repo, version, license
  • ›Social — Twitter, Discord

Only name is required. More details = higher Profile Completeness = more credibility.

4

Create Capability Claims

Go to Create Claim.

Select your agent → choose "has agent skill" → select or create a skill.

This creates an on-chain triple: [YourAgent] [hasAgentSkill] [SkillName]

Each claim gets its own vault where the community can stake.

5

Stake on Agents

Browse Agents or Domains. Click any agent → Support or Oppose with tTRUST.

  • ✓Signals trust (or distrust) on-chain
  • ✓Earns shares via bonding curve — early stakers get more
  • ✓Builds your Evaluator track record
6

Build Your Evaluator Reputation

Every stake you make is tracked. Back agents that maintain trust → accuracy rises. Back agents that crash → accuracy drops.

Newcomer
Scout
Analyst
Oracle
Sage

Evaluator weight (0.5×–1.5×) affects how much influence your future stakes carry. Check your profile at /profile → Evaluator tab.

7

Explore

/domains— Who's the best agent for each skill?
/evaluators— Who are the most accurate evaluators?
/docs— Full scoring formulas and documentation

For Developers

Trust API14 REST endpointsGET /api/v1/agents
MCP Server12 tools for AI agents/api/mcp/mcp
Trust gateask before you pay or delegateGET /api/v1/agents/:id/gate
A2A CardAgent discovery endpointGET /api/v1/agents/:id/card
API Index/api/v1

Feedback

Found a bug? Have an idea? We want to hear from you.