MCP server for checking code and terminal commands against organizational compliance policies
Server provides 4 tools with clear naming (verb_noun convention) and reasonable descriptions. However, schema completeness is inconsistent: check_compliance and check_command have rich input schemas with proper types and descriptions, but output schemas are entirely undocumented. Parameter descriptions are present but vary in quality. Error handling is basic. The server lacks tool annotations (readOnlyHint/destructiveHint) despite having READ_ONLY semantics. Tool composition is sound, tools are single-purpose and chainable. Overall: fair quality with notable gaps in output documentation and error guidance.
Check a terminal command (e.g. npm install, git push) against compliance policies before running it.
Check a code snippet or file content against the org's compliance policies for a given user role. Returns violations, their severity, and fix suggestions.
Fetch the compliance policies applicable to a given user role. Returns policy metadata, skills, and constraints to guide code generation.
Enable or disable autofix for this session. When enabled (default), the agent is told to self-correct blocked code using policy fix suggestions before asking the user. When disabled, every blocked check goes straight to the user.
Output schemas are entirely undocumented. Tools return compliance violations, severity levels, fix suggestions, and policy metadata, but no schema documentation is visible in the provided source. LLMs cannot predict the structure of responses, forcing them to guess at field names and infer types.
Tool annotations (readOnlyHint, destructiveHint, idempotentHint) are marked as false in the feature set, but the Risk field in tool definitions shows READ_ONLY semantics. These should be expressed as structured annotations in the tool schema, not metadata flags.
Parameter descriptions for optional policy-fetching parameters (policyUrl, authHeader) are generic (e.g., 'Remote URL', 'Auth header') and lack actionable constraints. No guidance on format, expected values, or mutual dependencies.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 61 | 2026-07-28+ | v2 |
Error handling is not documented. Tools accept authHeader and policyUrl but do not explain what errors may occur (invalid URL, 401 auth failure, 404 policy not found, network timeout). LLMs lack recovery guidance.
The userRole parameter is described as accepting comma-separated values ('payments-dev,ml-dev') to merge role chains, but this comma-separated format is not enforced in the schema. No validation or error guidance for malformed role strings.