MCP server trust scoring

AIM applies an 8-factor weighted algorithm to calculate trust scores for MCP servers, enabling security decisions about which servers your agents can connect to.

Trust factors

25%

Attestation weight

~50%

Initial score

80%+

Trusted threshold

How MCP trust scoring works

Every MCP server registered with AIM receives a trust score from 0-100 based on 8 weighted factors. This score determines whether agents can connect to the server and what level of monitoring is applied.

Trust score progression

Registered:~50% (pending verification)

1st attestation:~60% (+attestation consensus)

3+ attestations:~75% (consensus threshold met)

30+ days:~85% (+age and history bonus)

Fully trusted:90%+ (all factors optimized)

Trust score ranges

90%

Excellent

80-100: trusted

Agents can connect freely, minimal monitoring

70%

Good

60-79: acceptable

Agents can connect with standard monitoring

50%

Fair

40-59: caution

Connection requires policy approval, increased monitoring

30%

Poor

0-39: untrusted

Connections blocked by default, admin approval required

MCP vs agent trust scoring

MCP trust scoring uses its own 8-factor weighted approach, structured like the agent trust score but with MCP-specific factors:

Agent trust factors

• Verification Status (Ed25519)
• Uptime & Availability
• Action Success Rate
• Security Alerts

MCP trust factors

• Attestation Consensus
• Connection Health
• Capability Stability
• Security Posture

Viewing MCP trust scores in dashboard

Navigate to MCP Servers → [Server Name] to see:

• Overall trust score with color-coded indicator
• Factor-by-factor breakdown in radar chart
• Attestation progress and consensus status
• Score history over time
• Recommendations for improving trust score