Check a proposed agent action against operator's rules: permit, deny or escalate, with the clause.
Scored across 2 tools
check_action evaluates an action against a mandate, while validate_mandate validates a mandate's correctness. Their distinct purposes are clearly described, and the descriptions explicitly cross-reference each other, leaving no ambiguity about when to use which.
Both tool names follow a consistent verb_noun pattern (check_action, validate_mandate). The verbs differ appropriately (check vs. validate) based on the operation, but the structure is uniform and predictable.
The server's purpose is narrow—checking actions against mandates and validating mandates—so two tools cover the essential functionality. However, this is slightly below the typical 3–15 range, making it feel a bit thin even though both tools are necessary.
The server is stateless and focused on policy checking. The two tools cover the full lifecycle: validating a mandate before use and checking actions against it. No obvious gaps exist for its stated scope, as mandate management is presumably handled elsewhere.
Call this before your agent does something consequential (sends, pays, deletes, publishes, installs or changes access). Pass mandate (the operator's rules; the first matching clause decides) and action (kind, verb, target, optional counterparty and amount). Returns permit, deny or escalate with the governing clause, a reason, an action digest and a signed receipt; treat error as deny. Advisory: it cannot stop an action, and it keeps no request content. To check a new mandate first, call validate_mandate. Examples: inputSchema. Docs: https://hanria.ai/llms.txt
| Name | Required | Description | Default |
|---|---|---|---|
| action | Yes | A typed description of a proposed operation, submitted by an agent to a local HANRIA runtime. MUST NOT contain credentials, secrets, private keys, or tokens. | |
| mandate | Yes | An operator-authored statement of what an agent may do. Written by a person, read by an agent, evaluated locally. A mandate is advisory: it describes intended authority and supports a local check, but it does not enforce anything. Enforcement requires a component that exclusively holds the credentials and can refuse. |
| Name | Required | Description |
|---|---|---|
| clause | No | The clause that decided, when one did. |
| reason | Yes | |
| outcome | Yes | Treat error as deny. |
| receipt | No | Ed25519-signed receipt binding the outcome to this action and mandate for 15 minutes. Keys and the signed-bytes rule: https://check.hanria.ai/.well-known/hanria-receipt-keys.json |
| clause_note | No | |
| mandate_ref | No | |
| action_digest | No | SHA-256 of hanria-action-v1 followed by the action's canonical JSON. |
| schema_version | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Beyond the readOnly/idempotent annotations, it discloses the decision vocabulary (permit, deny, escalate), the returned artifacts (governing clause, reason, action digest, signed receipt), the error-handling rule ('treat error as deny'), and the crucial limitation that it is advisory and 'cannot stop an action' and keeps no request content. These are exactly the behaviors annotations cannot convey.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is front-loaded with purpose and triggering conditions, then moves to parameters, outputs, caveats, and the sibling pointer, with the two Examples/Docs pointers at the tail. It is dense but nearly every sentence carries weight; the parenthetical nesting makes parts slightly harder to scan than ideal.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a complex nested schema with an output schema present, the description supplies what the schema cannot: invocation timing, the advisory-only nature, error semantics, and sibling routing. An agent has everything needed to call it correctly and interpret results.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, which sets a baseline of 3, but the description adds real semantic value: it explains mandate carries the operator's rules and that 'the first matching clause decides,' and summarizes the action shape (kind, verb, target, optional counterparty and amount). This clarifies evaluation order beyond the schema's field-level docs.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states a specific verb and resource ('Check an agent action against a mandate') and frames the timing ('before your agent does something consequential'). It also distinguishes itself from the sibling by naming validate_mandate as the tool for checking a new mandate first. An agent can identify the tool's role without opening the schema.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It explicitly says when to call it ('before your agent does something consequential') and enumerates triggering actions (sends, pays, deletes, publishes, installs, changes access). It also routes to the alternative: call validate_mandate to check a new mandate first. Both the when and the alternative are stated outright.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Call this after writing or editing a mandate, before relying on it. Returns valid true or false with the reason: schema errors, clauses that can never run because an earlier clause already decides them, unbounded prefixes, credential material, and expiry. An invalid mandate makes every check_action call answer error. Once per mandate version is enough; there is no need to call it before every check_action. Free, and it keeps no request content. Docs: https://hanria.ai/llms.txt
| Name | Required | Description | Default |
|---|---|---|---|
| mandate | Yes | An operator-authored statement of what an agent may do. Written by a person, read by an agent, evaluated locally. A mandate is advisory: it describes intended authority and supports a local check, but it does not enforce anything. Enforcement requires a component that exclusively holds the credentials and can refuse. |
| Name | Required | Description |
|---|---|---|
| valid | Yes | |
| reason | Yes | |
| clauses | No | |
| expired | No | True when not_valid_after has passed; check_action then denies everything. |
| mandate_ref | No | |
| mandate_digest | No | SHA-256 of the mandate's canonical JSON; receipts carry the same value. |
| schema_version | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations cover the safety profile (readOnly, idempotent, closed-world), and the description adds non-structured traits: it is free, retains no request content, and enumerates the exact error classes it reports (schema errors, unreachable clauses, unbounded prefixes, credential material, expiry). It also explains the consequence of an invalid mandate on check_action, which is real behavioral context.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Front-loaded with the action and timing, then the error taxonomy, then frequency and cost. Dense but mostly purposeful; the docs link and the return-value mention are slightly redundant given the output schema exists, but nothing is bloated.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a validation tool with a rich nested schema and an output schema, the description covers what an agent needs: when to call, what it detects, the downstream consequence, and cost/privacy. Return format is already handled by the output schema.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Only one parameter exists and schema coverage is 100%, so the schema already documents the mandate object in depth (including the credential-material caveat and prefix-matching semantics). The description adds no parameter-level detail beyond what the schema provides, so the baseline 3 applies.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb and resource ('Check a mandate for errors') and clarifies it validates a mandate rather than an action, which cleanly separates it from the sibling check_action. An agent can identify the tool's role without opening the schema.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Gives explicit timing ('after writing or editing a mandate, before relying on it'), an exclusion ('no need to call it before every check_action'), and a frequency rule ('once per mandate version is enough'). This is unusually complete routing guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
check_action1 field changedOutput schema / (root)Previous value: -nullNew value: +{
+ "additionalProperties": false,
+ "properties": {
+ "action_digest": {
+ "description": "SHA-256 of hanria-action-v1 followed by the action's canonical JSON.",
+ "type": "string"
+ },
+ "clause": {
+ "description": "The clause that decided, when one did.",
+ "type": "string"
+ },
+ "clause_note": {
+ "type": "string"
+ },
+ "mandate_ref": {
+ "type": "string"
+ },
+ "outcome": {
+ "description": "Treat error as deny.",
+ "enum": [
+ "permit",
+ "deny",
+ "escalate",
+ "error"
+ ],
+ "type": "string"
+ },
+ "reason": {
+ "type": "string"
+ },
+ "receipt": {
+ "additionalProperties": false,
+ "description": "Ed25519-signed receipt binding the outcome to this action and mandate for 15 minutes. Keys and the signed-bytes rule: https://check.hanria.ai/.well-known/hanria-receipt-keys.json",
+ "properties": {
+ "action_digest": {
+ "type": "string"
+ },
+ "clause": {
+ "type": "string"
+ },
+ "issued_at": {
+ "format": "date-time",
+ "type": "string"
+ },
+ "kid": {
+ "type": "string"
+ },
+ "mandate_digest": {
+ "type": "string"
+ },
+ "mandate_ref": {
+ "type": "string"
+ },
+ "not_after": {
+ "format": "date-time",
+ "type": "string"
+ },
+ "outcome": {
+ "enum": [
+ "permit",
+ "deny",
+ "escalate"
+ ],
+ "type": "string"
+ },
+ "signature": {
+ "type": "string"
+ },
+ "type": {
+ "const": "hanria-receipt-v1",
+ "type": "string"
+ }
+ },
+ "required": [
+ "type",
+ "kid",
+ "outcome",
+ "action_digest",
+ "mandate_digest",
+ "issued_at",
+ "not_after",
+ "signature"
+ ],
+ "type": "object"
+ },
+ "schema_version": {
+ "const": "0.1-draft",
+ "type": "string"
+ }
+ },
+ "required": [
+ "schema_version",
+ "outcome",
+ "reason"
+ ],
+ "type": "object"
+}validate_mandate1 field changedOutput schema / (root)Previous value: -nullNew value: +{
+ "additionalProperties": false,
+ "properties": {
+ "clauses": {
+ "type": "integer"
+ },
+ "expired": {
+ "description": "True when not_valid_after has passed; check_action then denies everything.",
+ "type": "boolean"
+ },
+ "mandate_digest": {
+ "description": "SHA-256 of the mandate's canonical JSON; receipts carry the same value.",
+ "type": "string"
+ },
+ "mandate_ref": {
+ "type": "string"
+ },
+ "reason": {
+ "type": "string"
+ },
+ "schema_version": {
+ "const": "0.1-draft",
+ "type": "string"
+ },
+ "valid": {
+ "type": "boolean"
+ }
+ },
+ "required": [
+ "schema_version",
+ "valid",
+ "reason"
+ ],
+ "type": "object"
+}check_action1 field changedInput schema / examplesPrevious value: -[
- {
- "action": {
- "justification": "Read the permitted file.",
- "operation": {
- "kind": "file",
- "target": "/tmp/example",
- "verb": "read"
- },
- "requested_by": {
- "agent": "example-agent"
- },
- "schema_version": "0.1-draft"
- },
- "mandate": {
- "clauses": [
- {
- "effect": "permit",
- "id": "permit-file",
- "match": {
- "kind": [
- "file"
- ]
- }
- }
- ],
- "default": "deny",
- "mandate_id": "minimal-permit",
- "purpose": "Permit one file action.",
- "schema_version": "0.2-draft"
- }
- }
-]New value: +[
+ {
+ "action": {
+ "justification": "Read the permitted file.",
+ "operation": {
+ "kind": "file",
+ "target": "/tmp/example",
+ "verb": "read"
+ },
+ "requested_by": {
+ "agent": "example-agent"
+ },
+ "schema_version": "0.1-draft"
+ },
+ "mandate": {
+ "clauses": [
+ {
+ "effect": "permit",
+ "id": "permit-file",
+ "match": {
+ "kind": [
+ "file"
+ ]
+ }
+ }
+ ],
+ "default": "deny",
+ "mandate_id": "minimal-permit",
+ "purpose": "Permit one file action.",
+ "schema_version": "0.2-draft"
+ }
+ },
+ {
+ "action": {
+ "justification": "Pay the invoiced supplier.",
+ "operation": {
+ "amount": {
+ "currency": "USDC",
+ "value": "25"
+ },
+ "counterparty": "0x2222222222222222222222222222222222222222",
+ "kind": "transaction",
+ "target": "eip155:11155111/0x1111111111111111111111111111111111111111/transfer",
+ "verb": "transfer"
+ },
+ "requested_by": {
+ "agent": "example-agent"
+ },
+ "schema_version": "0.1-draft"
+ },
+ "mandate": {
+ "clauses": [
+ {
+ "effect": "permit",
+ "id": "small-transfers",
+ "match": {
+ "counterparty": [
+ "0x2222222222222222222222222222222222222222"
+ ],
+ "kind": [
+ "transaction"
+ ],
+ "max_amount": {
+ "currency": "USDC",
+ "value": "100"
+ },
+ "target_prefix": [
+ "eip155:11155111/0x1111111111111111111111111111111111111111/"
+ ],
+ "verb": [
+ "transfer"
+ ]
+ }
+ }
+ ],
+ "default": "escalate",
+ "mandate_id": "example-payments",
+ "purpose": "Permit small transfers to one named recipient on a test network.",
+ "schema_version": "0.2-draft"
+ }
+ }
+]validate_mandatecheck_actionCheck an agent action against its operator's mandate before acting: permit, deny or escalate.
Deterministic allow/require_approval/deny verdicts for agent actions, before they happen.
Git-native policy layer for AI agents: check_action verdicts against rules approved via PR.
The system of record for AI agent authority: playbooks, routed policy questions, reusable rules.
Add one secure layer between your agents and this server.
