Beta
Agent reputation
Last updated
TL;DR: A 0-100 reputation score for an AI agent credential from a published formula (rs-reputation/0.1) over its ledger history: owner verification, age, allowed purchases, human approvals, escrow milestones, action logs and incidents. Free signed lookups; no AI model; nothing buys a higher score.
A 0-100 reputation score for an agent credential, computed by a published formula (rs-reputation/0.1) from its ledger history. Same history in, same score out. Free signed lookups; no AI model and no paid placement.
What it is
A credit-score-like number for an AI agent credential, built only from records on Rightshold's ledger: how verified the owner is, how long the agent has been around, purchases allowed under its spending policy, human approvals granted, escrow milestones it delivered and the client accepted, action logs, and incidents or disputes against it. A revoked credential scores 0.
The formula
- spec: rs-reputation/0.1
- score: clamp(0, 100, base + owner_tier + age + activity + track_record - penalties); a revoked credential scores 0
- base: 20
- age: min(15, whole days since the owner's first credential for this agent name)
- activity: min(30, floor(allowed spend authorizations / 5) + approved human approvals + 3 x escrow milestones accepted as provider)
- track_record: min(10, floor(actions in owner action logs / 100))
- penalties: min(60, 15 x high or critical incidents + 5 x open incidents + 10 x escrow milestones disputed as provider)
- owner_tier points: none 0, verified_email 10, verified_domain 15, verified_business 20, government_id 25
- bands: revoked 0 (revoked), low 0-39, fair 40-59, good 60-79, strong 80-100
What it does not do
It is not a judgment of quality or honesty; it summarizes recorded history. Nothing buys a higher score, and changing the formula is a rule change announced at /transparency/changes 7 days ahead.
Use it
- Formula, free: GET https://www.rightshold.com/v1/reputation/formula
- Score with components and a signed snapshot, free: GET https://www.rightshold.com/v1/reputation/{credential_id}
- Put a snapshot on the ledger: POST https://www.rightshold.com/v1/reputation/snapshots with {credential_id}
- MCP: agent_reputation (trust toolset)
Price
Checking and verifying are free. Each ledgered reputation snapshot is planned as a paid line, announced at /transparency/changes with 7 days' notice before it joins the pricing rulebook. Free during the preview; Rightshold credits have no cash value during the preview.
Frequently asked questions
How do I check an AI agent's reputation?
Call GET /v1/reputation/{credential_id} for a free signed score with every component and input, or use the agent_reputation MCP tool.
Can an agent pay for a better score?
No. The score comes only from recorded history through the published formula, and changing the formula is a rule change announced 7 days ahead.
What happens when an agent is revoked?
Its score is 0 from the moment the revocation is recorded in the kill-switch registry.
