Deny-by-default authority leases for agents wielding real power.
Scored across 7 tools
Each tool targets a distinct phase of the covenant lifecycle: introspection (describe_agent), issuance (grant_covenant), enumeration (list_covenants), termination (revoke_covenant), state inspection (covenant_status), enforcement (check_act), and audit verification (verify_audit). There is no realistic risk of confusing check_act with covenant_status or grant_covenant, since their descriptions delineate side effects and read-only behavior clearly.
All names use consistent snake_case, and six of seven follow a verb_noun pattern (grant_covenant, revoke_covenant, list_covenants, check_act, describe_agent, verify_audit). The outlier is covenant_status, which is noun-first rather than a verb-led command, but it remains immediately readable and semantically unambiguous.
Seven tools is well within the ideal range and each earns its place: every major operation in the covenant domain has exactly one corresponding tool. There are no redundant aliases or filler operations, and the surface stays small enough for an agent to reason about holistically.
The core lifecycle is covered end to end: grant, list, status, revoke, enforce via check_act, and verify the audit chain. Minor gaps exist, such as no dedicated get_covenant for fetching a single covenant's full detail (likely folded into list_covenants) and no operation to amend scopes or extend a budget/expiry without revoking and re-granting.
Check (and if allowed, record) a proposed act against a covenant. Deny-by-default. Idempotent on act_id — retries never double-consume budget. Every check lands on the audit chain.
Inputs: scope (required), act_id (required), covenant_id (required), amount_minor. Returns: result
Current state, consumed/remaining budget, and check count.
Inputs: covenant_id (required). Returns: result
Describe this agent: name, version, purpose, every action it supports and the inputs each one takes. Read-only and free; call it first to learn how to use this server.
Inputs: no named parameters. Returns: result
Grant an agent a covenant: an explicit lease of authority. scopes support wildcards ('payments.', ''); budget_minor is the total spend ceiling in minor units; expires_at is ISO-8601. Returns the covenant_id.
Inputs: scopes (required), agent_id (required), principal (required), expires_at (required), +1 more. Returns: result
List covenants, optionally filtered by state (ACTIVE|REVOKED|EXPIRED) and/or the bound agent.
Inputs: state, agent_id. Returns: result
Revoke a covenant immediately and terminally. All subsequent checks deny.
Inputs: covenant_id (required), reason. Returns: result
Verify the tamper-evident audit chain of allowed/denied acts.
Inputs: covenant_id (required). Returns: result
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Machine-verifiable authorization checks for autonomous AI agents.
Bounded authorization and independently verifiable execution evidence for agents.
Pre-execution policy gate for consequential agent actions with durable trust receipts.
Identity and authorization in one system: allow, deny, the reason, and a record.
Add one secure layer between your agents and this server.
