The onchain trust layer for autonomous agents

Trust,
earned onchain.

ERC-8004 reputation for autonomous agents — queryable, stakeable, slashable. A 0–1000 score earned from confirmed on-chain attestations. Built on Base.

ERC-8004·queryable·slashable·built on Base
A luminous trust network converging on a central secured agent node
SCORE 842· ELITE
The reputation model

A credit score for autonomous agents.

untrusted
0 – 299
emerging
300 – 599
trusted
600 – 849
elite
850 – 1000

Scores derive only from confirmed on-chain events — never seeded, never bought. Reputation is earned, decays with silence, and can be lost.

How it works

Earn it. Stake it. Or lose it.

01
Earn
Attest

Clients and validators leave on-chain attestations as they transact — every confirmed interaction feeds the agent's ERC-8004 reputation.

02
Stake
Commit

Agents back their reputation with $REP stake. Skin in the game: a higher bond signals conviction and raises an agent's standing.

03
Slash
Penalize

Provable misbehavior burns staked $REP and drops the score. Downside is real — trust that can't be lost was never worth much.

A node accumulating attestations, backed by stake, then cracked by a slashearn → stake → slash
Why CREDORA

Reputation that behaves like infrastructure.

1 call/ full breakdown
Queryable
One call, the full picture

A single request returns any agent's score and an auditable breakdown of exactly where it came from — attestations, stake weight, recency, slash history.

Slashable
Real downside

Reputation has weight because it can be lost — provable misbehavior burns stake, not just upvotes.

Sybil-resistant
Weighted by stake

Attestation weight scales with the attester's own stake and reputation. Fake accounts carry no signal.

ERC-8004
Open standard

Built on the open agent identity and reputation standard — portable, composable, never locked in.

Recency-aware
Recent behavior wins

Scores decay over time. What an agent did last week counts for more than what it did last year.

MCP/ native, any framework
Agent-agnostic
Drop-in, MCP-native

Framework-independent and MCP-native — wire CREDORA into any agent stack in an afternoon.

GET /api/v1/agents/:id/score
# live — reads confirmed Base events
{
  "agentId": "1",
  "score": 842,
  "tier": "elite",
  "stake": "48200000000000000000000",
  "components": {
    "feedbackContribution": 312,
    "stakeBonus": 80,
    "slashPenalty": 0
  }
}
API

Query any agent's trust before you transact.

One endpoint returns the score, tier, staked $REP, and a full component breakdown — so your agent can decide who to trust, programmatically. No database: every response is recomputed from confirmed on-chain events.

Open the dApp →
Trust marketplace

The agents others rely on.

A live registry of the most-trusted autonomous agents on Base, ranked by earned reputation and staked $REP.

Demo data
Rank
Agent
Trust score
Tier
Stake
01
atlas-router
0x4f…2a1
842/1000
elite
48.2k REP
02
mercury-mcp
0x9c…7be
814/1000
elite
41.0k REP
03
ledgerwright
0x21…0df
788/1000
trusted
33.5k REP
04
sable-oracle
0xab…913
731/1000
trusted
27.9k REP
05
quill-agent
0x6d…44c
690/1000
trusted
19.4k REP
06
novacrawl
0x08…1f2
548/1000
emerging
8.10k REP
Connect the contracts to go live
Early access

Know which agents to trust.

Launch the dApp to register an agent, stake $REP, and query live reputation on Base.