codebase-doctor
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| audit_codebaseA | Run the full built-in Codebase Doctor audit on a repository and return the evidence-backed report. Read-only and offline by default; it never enables validation commands (--run-checks) or live database access (--with-database). |
| describe_capabilitiesA | Describe this MCP server: available tools, the nine audit domains in the domainCoverage inventory, the Doctor capability vocabulary, and the permissions this server never grants. |
| verify_changesA | Verify that findings from a prior schema-1 JSON report are repaired. Runs a fresh read-only audit, compares fingerprints, and reports each baseline finding as resolved, unchanged, or unresolved. Absence under incomplete coverage is unresolved, never resolved. Read-only and offline by default; it never enables validation commands or live database access. |
| explain_findingA | Return the full evidence, remediation guidance, and verification command for one finding, selected by fingerprint or rule id. Read-only and offline. |
| review_changesA | Review changed code with a PR-ready verdict. Always audits changed scope and narrows findings to added diff lines, returning APPROVE, COMMENT, or REQUEST_CHANGES. Findings on unchanged lines are omitted, never resolved. Read-only and offline by default; it never enables validation commands or live database access. |
| verify_intentA | Verify declared intent against audit evidence: claims (rule-absent, rule-present, score-at-least, coverage-complete) are checked as verified, violated, or undecided. Intent is declared JSON or a fenced ```intent block; prose is never interpreted. Undecided means the claim depends on coverage that did not complete and is never counted as verified. Read-only and offline. |
| build_receiptA | Run a read-only audit and return a portable coverage receipt: what was checked, what was not, the score, and finding fingerprints, with a SHA-256 digest over the canonical body. Receipts never contain secret values or source text. Read-only and offline. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 7 tools
Most tools target clearly distinct audit-lifecycle operations, and descriptions distinguish full audit, changed-scope review, repair verification, and intent verification. The main overlap is audit_codebase vs build_receipt, both of which run a read-only audit, though their outputs (full report vs coverage receipt) are sufficiently differentiated.
All seven tool names follow a consistent snake_case verb_noun pattern. The verbs are varied but predictable, and there are no mixed conventions or naming collisions.
Seven tools is well within the appropriate range for a focused audit/verification server. Each tool maps to a distinct operation in the audit lifecycle, and there is no apparent filler.
The surface covers auditing, receipt generation, changed-code review, repair verification, finding explanation, intent verification, and capability discovery. Minor gaps exist around dedicated finding enumeration/filtering and receipt comparison or verification, but agents can work around these using the existing tools.