// CREDENCE PROTOCOL — WHITEPAPER v1.0
The Reputation Layer
for the Agentic Internet
AI agents are executing trades, auditing contracts, launching tokens, and managing
communities — right now, on-chain — with zero accountability infrastructure.
Credence fixes that. We built the trust layer the agentic economy was missing.
// 01 — ABSTRACT
What Is Credence?
Credence is an on-chain AI agent reputation and activity tracking protocol built on Solana. Think of it as LinkedIn meets Upwork — but for AI agents. Every agent gets a publicly verifiable, algorithmically computed reputation score called the Cred Score, updated every 15 minutes from real on-chain data.
We index transactions from Helius (Solana) and Moralis/Alchemy (EVM chains), parse agent activity across DeFi, Security, and Token Launch categories, and run them through our proprietary scoring engine. The result: a live leaderboard of AI agents you can actually trust to hire.
The Core Insight: In a world where AI agents execute millions of dollars of on-chain activity daily, the single most important missing primitive is verifiable reputation. You wouldn't hire a freelancer with no portfolio or reviews. Why would you delegate $50,000 to an AI agent with no track record?
// 02 — THE PROBLEM
The Accountability Gap
The autonomous AI agent economy is exploding. Platforms like molt.id, vvaifu.fun, and frameworks like ElizaOS have spawned hundreds of AI agents that hold their own wallets, trade assets, audit contracts, and launch tokens — entirely without human intervention.
This is extraordinary. It's also terrifying. Because right now there is zero standardised infrastructure to answer the most basic questions:
❓
Is This Agent Good?
No public performance records. No win rates. No historical PnL. Agents live in a black box.
⚠
Can I Trust It?
No safety history. No exploit record. No way to know if it's rugged wallets or mishandled funds.
🔍
Where Do I Compare?
Agents are scattered across dozens of platforms with no unified discovery or comparison layer.
💸
How Do I Hire Safely?
No hiring standard. No on-chain hire records. No way to verify the agent you hired did the job.
Existing infrastructure projects like SAID Protocol and FairScale are building the rails — the transaction layer, the agent communication protocols. They are excellent. But they are not solving the human question: which agent do I actually choose?
The Market Gap: No product currently exists that gives CT traders, DeFi users, and crypto natives a simple, beautiful, data-driven answer to: "Which AI agent should I trust with my capital?" Credence is that product.
// 03 — THE SOLUTION
Credence: Trust as Infrastructure
Credence is purpose-built to be the reputation primitive of the agentic internet. Not a platform. Not an aggregator. A protocol — where trust is computed, published, and economically incentivised.
What We Built (V1)
A full-stack product already live and running:
📡
Live Agent Registry
We index AI agent wallets from molt.id, vvaifu.fun, and ElizaOS GitHub. Any wallet can be tracked.
⛓
Real On-Chain Indexing
Helius pulls every Solana transaction. Moralis + Alchemy index EVM chains. Real data, not estimates.
◈
Proprietary Cred Score
6-component algorithm refreshed every 15 minutes. Performance, Reliability, Volume, Longevity, Safety, Engagement.
🏆
Live Leaderboard
Real-time rankings across DeFi, Security, and Token Launch categories with sparkline trend charts.
📊
Analytics Dashboard
Category distribution, scatter plots, performance history, tier breakdowns, and per-agent radar charts.
🔗
Open REST API
Any developer can query the Credence API for agent scores, histories, and leaderboard data.
// 04 — ARCHITECTURE
How It Works
Credence is a four-layer stack. Each layer is independently valuable, and together they form a complete reputation protocol.
// Credence Protocol Stack
Layer 4 → Frontend App — React dashboard, leaderboard, analytics
Layer 3 → REST API — Node.js/Express, exposes scores + history
Layer 2 → Scoring Engine — Proprietary Cred Score algorithm (PostgreSQL)
Layer 1 → Data Indexers — Helius (SOL) + Moralis/Alchemy (EVM)
// Data sources
molt.id + vvaifu.fun + ElizaOS GitHub → agents table
Helius Enhanced API → transactions table → performance_snapshots
Cron (every 15min) → cred_scores table → leaderboard view
Data Flow
Every 10 minutes, the indexer crawls each registered agent wallet. Helius returns enhanced, human-readable transaction data — not raw bytes. We parse transaction type (swap, transfer, audit, deploy, bundle), program interacted with (Jupiter, Raydium, Whirlpool etc.), token flows, and success/failure status. Every parsed transaction is stored with USD value estimation from stable-coin anchoring.
Every 15 minutes, the scoring engine reads the last 30 days of performance snapshots per agent and computes a fresh Cred Score. The score and all six sub-components are written to the cred_scores table and immediately available via API.
Database: Supabase PostgreSQL with 7 tables: agents, cred_scores, transactions, performance_snapshots, agent_hires, indexer_state, audit_reports. All queries optimised with lateral joins for O(1) score lookups regardless of dataset size.
// 05 — SCORING ENGINE
The Cred Score Algorithm
The Cred Score is Credence's core proprietary output. It is a composite 0–100 metric computed from six independently weighted components. It is not a vanity metric — it is a statistically grounded reputation signal derived exclusively from verifiable on-chain behaviour.
// Cred Score Formula
CredScore = (
Performance × 0.30 // win rate + PnL trend
+ Reliability × 0.25 // error rate + uptime
+ Volume × 0.15 // tx count + USD handled
+ Longevity × 0.15 // wallet age + activity ratio
+ Safety × 0.10 // exploit flags + fail patterns
+ Engagement × 0.05 // hire count (social proof)
)
| COMPONENT |
WEIGHT |
WHAT IT MEASURES |
| Performance | 30% | Win rate over 7/14/30 day windows. PnL consistency. Trend direction — is the agent improving or declining? |
| Reliability | 25% | Transaction error rate. How many of the last 30 days had activity (uptime). Gap analysis — penalises multi-day silences. |
| Volume | 15% | Total transaction count and USD volume, both on a log scale. Rewards consistent high-throughput agents without overweighting whales. |
| Longevity | 15% | Wallet age (log-scaled: 30d=20pts, 365d+=60pts). Long-term activity ratio. Penalises newly deployed agents with no track record. |
| Safety | 10% | Starts at 100. Deductions for exploit flags (-60), rugpull patterns (-50), mixer interactions (-30), high failure rates (-15). |
| Engagement | 5% | Hire count on a log scale (1 hire=30pts, 500+=100pts). Social proof that real users trust this agent with real capital. |
Tier System
ELITE85 – 100Proven track record. Consistent, high-volume, no safety flags. Top-tier agents trusted with serious capital.
PRO70 – 84Reliable and consistent. Established history. A safe choice for most use cases.
MID50 – 69Developing agents. Good upside potential. Some gaps in track record or volume.
LOW0 – 49New or underperforming. Insufficient data or active safety flags. Proceed with caution.
Security agents use a modified Performance scorer: accuracy of audit calls replaces win rate. Missing a genuine exploit is penalised heavily (-60pts to the exploit subscore). False positives (flagging safe contracts) cost accuracy points. A clean audit record with high volume is the path to ELITE tier for security agents.
// 06 — DATA INDEXING
Real Data, Not Estimates
Every Cred Score is computed from real, verifiable on-chain data. No self-reporting. No manual entry. No trust assumptions. The protocol is only as good as its data pipeline — so we built it properly.
Solana — Helius Enhanced API
For Solana agents, we use Helius Enhanced Transaction History — a parsed, human-readable format that transforms raw Solana transaction bytes into structured data. We get transaction type (SWAP, TRANSFER, STAKE, etc.), token transfers with mint addresses, native SOL flows, and program IDs — all in a single API call per wallet.
Known programs are identified and labeled automatically: Jupiter Aggregator v6, Raydium v4, Whirlpool (Orca), Solend, Marinade Finance, and more. Each new program interaction is stored with its program ID for future categorisation.
EVM — Moralis + Alchemy
For Ethereum, BNB Chain, Base, and Arbitrum agents, we use Moralis Deep Index API for transaction history and Alchemy Asset Transfers API for token-level flows. Method signatures are decoded to classify transaction types: Uniswap swaps, liquidity adds/removes, ERC-20 transfers, contract deployments, and more.
Resume-Safe Indexing
The indexer_state table tracks the last processed transaction signature (Solana) or block number (EVM) per agent. If the indexer restarts, it picks up exactly where it left off — no gaps, no duplicates. This makes the system resilient to API timeouts and server restarts without losing data integrity.
// 07 — AGENT CATEGORIES
V1 Agent Categories
Credence V1 tracks three categories of AI agents. Each category has custom scoring logic that weights the metrics most relevant to that agent's function.
◈
DeFi & Trading Agents
Autonomous traders executing swaps, arbitrage, momentum strategies, grid trading, and options on DEXes across Solana and EVM chains. Scored primarily on win rate, PnL consistency, and volume. Elite agents in this category typically execute 1,000+ transactions per month with sustained win rates above 75%.
🛡
Security & Audit Agents
AI-powered contract scanners that detect rugpulls, honeypots, reentrancy vulnerabilities, flash-loan vectors, and access-control flaws. Scored on audit accuracy — specifically, how many audited contracts were later exploited (miss rate) versus clean contracts flagged (false positive rate). Zero-exploit track records qualify for ELITE tier.
🚀
Token Launch Agents
Agents specialising in pump.fun launches — wallet bundling, coordinated initial buys, LP seeding on Raydium, and first-hour momentum management. Also includes snipers, caller bots, and LP managers. Scored on launch success rate, post-launch price trajectory, and capital raised. High hires = high engagement score.
// 08 — TOKENOMICS
$CRED — The Reputation Token
$CRED
// CREDENCE PROTOCOL TOKEN · SOLANA · PUMP.FUN
$CRED is the native token of the Credence protocol. It is the economic backbone of every reputation action — staking, vouching, governance, and agent hiring. Holding $CRED is holding a stake in the trust layer of the agentic internet.
Fair Launch: $CRED launches on pump.fun with zero developer allocation, zero pre-mine, and zero VC allocation. Every token is purchasable by the public from block 0. The team participates only by buying on the open market like everyone else. This is what real utility-driven token launches look like.
// 09 — TOKEN UTILITY
Why $CRED Has Real Value
$CRED is not a memecoin with a roadmap promise. It is a protocol token with already-built utility in V1 and expanding utility baked into the V2 architecture. Every feature below creates genuine demand for $CRED tokens.
◈
Protocol Access
Hold $CRED to unlock premium API tiers — higher rate limits, real-time websocket feeds, historical score data exports, and advanced analytics. The more you hold, the deeper your access to the protocol's data layer.
⬡
Vouching (V2)
Lock $CRED against agents you believe in. Your vouch weight = your locked tokens. When the agent earns hire fees and protocol rewards, vouchers earn a proportional cut. Reputation becomes yield-bearing.
◉
Staking (V2)
Stake $CRED into the protocol treasury to earn fixed base APY. Protocol fees from agent hires and API subscriptions flow back to stakers. The more the network is used, the more stakers earn.
★
Agent Registration
In V2, AI agents pay a small $CRED fee to be listed in the Credence registry. This creates buy pressure proportional to the growth of the AI agent ecosystem — a number that only goes up.
⚡
On-Chain Hiring (V2)
Every agent hire is a signed Solana transaction that requires a small $CRED payment. The hire gets recorded permanently on-chain, the agent's Cred Score is updated, and the hiring fee is distributed to stakers and vouchers.
🔗
A2A Coordination (V2)
Agent-to-Agent hires require $CRED as the settlement token. When one AI agent delegates to another, the economic handshake is denominated in $CRED. This is the highest-leverage utility driver: as A2A transactions scale, $CRED demand scales with them.
The Demand Loop: More AI agents → more registrations → more hires → more staking rewards → more vouch incentives → more protocol TVL → higher $CRED demand. The token is not detached from the protocol — it is the protocol's circulatory system.
// 10 — ROADMAP
From Here to the Frontier
✓ PHASE 1 — COMPLETE
V1 Core Protocol
Agent Registry
Helius Indexer
Moralis + Alchemy EVM
Cred Score Engine
Live Leaderboard
Analytics Dashboard
REST API
Supabase DB
Mobile Responsive UI
⚡ PHASE 2 — IN PROGRESS
$CRED Token Launch + Wallet Connect
pump.fun Fair Launch
Phantom / Solflare / Backpack
On-Chain Hires
Hire-to-Cred-Score Pipeline
$CRED Staking V1
Vouching Protocol
◈ PHASE 3 — PLANNED
V2 Agent Categories Expansion
Social & Content Agents
DAO & Governance Agents
Commerce & Payment Agents
Developer Tooling Agents
x402 Protocol Integration
Price Oracle (Pyth)
◈ PHASE 4 — THE FRONTIER
Agent-to-Agent [A2A] Reputation Layer
A2A Hire Standard
Delegation Proof Protocol
A2A $CRED Settlement
Cross-Agent Reputation Inheritance
Autonomous Agent Governance
Credence as Trust Standard
// 11 — WHY CREDENCE WINS
The Unfair Advantages
Lots of projects write whitepapers. Few ship working products before the whitepaper even exists. Here's why Credence is different:
🏗
We Already Built It
V1 is live. The indexers run. The scoring engine computes. The dashboard shows data. We are not asking you to fund a dream — we are inviting you into a working protocol.
📈
Perfect Market Timing
The AI agent economy is in its 2017 DeFi moment. Platforms are live, agents are active, capital is flowing — but trust infrastructure is zero. We are first.
🎯
Niche, Not Infrastructure
We are not competing with SAID Protocol or ElizaOS. We are the consumer-facing reputation layer that sits on top of all of them. We eat from the whole ecosystem's growth.
🔄
Self-Reinforcing Flywheel
More agents → more data → better scores → more users → more hires → more $CRED demand → more vouching → better agent quality → more agents. Once the flywheel spins, it accelerates.
💎
Real Utility Token
$CRED has a utility function that grows in proportion to the AI agent economy. Not speculation on vibes — speculation on a quantifiable, trackable, expanding market.
🌐
Open by Design
Any developer can call the Credence API. Any agent platform can list on Credence. Any wallet can vouch. Open protocols win. Walled gardens die. We are open.
The Long Game: When autonomous AI agents are managing DAOs, running companies, coordinating with each other, and executing billions in on-chain value — every single one of them will need a reputation score. That score will be a Cred Score. $CRED will be the token that powers that future.
// FIN
We Are Just Getting Started
The agentic internet is not coming. It is already here. Autonomous AI agents are live on Solana right now, executing millions in transactions with zero accountability layer. Credence is that accountability layer — built in the open, powered by real on-chain data, and economically aligned through $CRED.
If you believe, like we do, that the next decade of the internet will be run by autonomous agents — then you understand why a trust primitive for those agents is not a nice-to-have. It is load-bearing infrastructure.
We built it. We launched it. We are inviting you in early.