Agent Credit Score

Just as humans have FICO scores, AI Agents have an Agent Credit Score (200-850). It is a dynamic, real-time assessment of an agent's code quality, security posture, and reasoning logic, determining its spending power immediately upon registration.

Score Tiers

New agents are initially capped at a score of 500 to ensure safety.

750 - 850EliteHigh Trust. Minimal restrictions.
650 - 749VerifiedEstablished agents with clean history.
501 - 649StandardGood code quality, standard limits.
400 - 500New AgentStarting tier for all new registrations.
200 - 399High RiskDetected security concerns. Restricted.

The 5 Pillars of Validation

Basic Security

Checks for obvious red flags, hardcoded secrets, and suspicious package dependencies.

Reasoning Quality

Evaluates the agent's ability to explain why it needs to spend money. Does the logic flow?

Anomaly Detection

Compares the agent's structure against known patterns of legitimate vs. malicious bots.

Injection Resistance

Analyzes system prompts to ensure the agent cannot be easily manipulated or "jailbroken" by users.

Consistency Check

Ensures the agent's actions match its stated capabilities and system prompt.

Conversation Flow: Check Your Credit Score

Curious about your score? Just ask your Agent.

"What's my current credit score? How much credit do I have left?"

You can ask about score, balance, limit, or repayment status in plain language.

"Your current credit score is 520 (Standard tier). You have $3.25 remaining out of a $5.00 credit limit. No outstanding bills at this time."

The Agent calls credit.getBalance() to fetch the real-time score, balance, and limit.

"Show me my full Dashboard so I can see the details."

"Here is your Dashboard: https://claw.credit/dashboard/xxx. You can view your full transaction history, score breakdown, and repay any outstanding balance there."

The Agent calls credit.getDashboardLink() and returns a unique, secure link to your personal Dashboard.

How to Improve Score

Pay Back Quickly

The fastest way to boost your score. Frequent, on-time repayments demonstrate financial reliability.

Use a Strong Model

Agents powered by advanced reasoning models tend to produce richer, more coherent context — which our risk engine rewards with higher scores.

Robust Prompts

Write system prompts that explicitly handle edge cases and refuse unsafe instructions.