The first covenant-governed trust protocol for AI agents. Screen → Attest → Certify → Verify → Reputation → Revoke. Open-source SDK. MIT license.
Scored across 4 tools
Each tool targets a distinct governance function: budget enforcement, identity registration, legal screening, and approval requests. Boundaries are mostly clear, though lawful_check and request_approval could overlap in risk-handling contexts.
Three tools use a noun_verb pattern (budget_enforce, identity_register, lawful_check) while request_approval uses verb_noun. All are snake_case, but the verb placement is inconsistent.
Four tools is a well-scoped set for a minimal governance surface, with each tool covering a distinct concern. No obvious bloat or missing tool count.
The surface lacks key lifecycle operations: no way to set budget limits, check approval status, or update/delete identities. These gaps will hinder agent workflows that require feedback or configuration.
Check if an agent spending amount exceeds its monthly limit.
Inputs: amount (required), agent_id (required).
Register or retrieve an agent persistent identity.
Inputs: agent_id (required), role.
Screen a task for unlawful-adjacent elements.
Returns tier (CLEAR/FLAGGED/BLOCKED), categories, reasoning, and disclaimer.
Inputs: agent_id (required), description (required).
Request human approval for a risky action.
Returns an approval_id. The action stays pending until a human approves or denies.
Inputs: action (required), reason (required), agent_id (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Add one secure layer between your agents and this server.