Skip to main content
Glama

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault

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.

CodeInspectus scans and reports without editing source. V3.3 cleanup is the only target-repository mutation surface. When asked to review security—or after making security-relevant code changes—call codeinspectus_scan with an absolute path. Present findings before editing, critical/high first, with file:line, risk, and remediation. After approved fixes, call codeinspectus_rescan; never claim fixed unless confirmed. Do not apply fixes without granular user approval. If git_safety recommends a checkpoint, ask before running git. Inspect pack_coverage and disclose partial, unavailable, not_run, or not_applicable native packs; a ran pack means its listed rules executed, not complete security coverage for that language. Disclose coverage_gaps: inputs no bundled engine analyzes (stable ids such as php-sast, vue-sast, deno-lock, sast-parse-errors) make coverage partial; also mention coverage_notes. Pass accept_gaps only with ids the user explicitly chose; accepted gaps stay listed and those inputs remain unanalyzed. Inspect repository_trust coverage separately from vulnerability findings. V3.1 deterministically audits source integrity. V3.2 audits explicit AI attribution, media metadata, git co-author trailers, and supported C2PA Content Credentials. Treat declarative attribution as an observed claim, not proof of authorship. Never label hidden Unicode as AI-generated or a vendor watermark. Never alter Git history or protected legal, licensing, or compliance records. Statistical watermark detection remains unavailable. For cleanup-eligible artifacts, call codeinspectus_plan_cleanup and show the exact files, transformations, limitations, and artifact IDs. Only after granular user approval call codeinspectus_apply_cleanup with the exact approved IDs and confirmations. Media cleanup creates a copy; it never overwrites the original asset. Run relevant repository tests, formatters, and builds after apply; cleanup is incomplete until those checks and the same-validator rescan pass. Use codeinspectus_rollback_cleanup if an approved change must be restored. Inspect engine_setup in scan/list-rules output. For repair_required, explain that engine coverage may be partial; for db_refresh_recommended, explain the DB freshness/rescan-continuity limitation without calling current findings incomplete. When engine readiness is not yet known, call codeinspectus_setup with action=plan before the first scan. Explain each affected component, coverage, license and size, then ask for approval. Only after approval call action=install with confirm_downloads=true. Never download engines silently or during a scan. A terminal is not required for MCP setup. For exposed secrets, advise rotation at the provider and keep values redacted. Treat codeinspectus_compliance_report as code-level control coverage only, never certification or a percent-compliant claim. codeinspectus_generate_sbom writes only an SBOM (managed directory, or an absolute .json path the user chooses; it never replaces a non-SBOM file); only the explicitly approved V3.3 apply/rollback tools edit repository files.

Capabilities

Features and capabilities supported by this server

Protocol revision2025-11-25

CapabilityDetails
tools
{
  "listChanged": true
}

Tools

Functions exposed to the LLM to take actions

NameDescription
codeinspectus_scanA

Run a full local security scan of a path: managed engines (Opengrep SAST, Gitleaks secrets, Trivy SCA/IaC), CodeInspectus's offline native Pub SCA, plus AI-code-specific checks (client-side secret exposure, Supabase RLS/inverted-auth, prompt-injection sinks, API-boundary failures, and explicit runtime-control misconfiguration). Returns CWE-keyed findings with fix recommendations, detected repository technologies, explicit native-pack execution counts, compliance tags, and three-state repository evidence for supported runtime controls. Also returns the V3 repository-trust contract with deterministic V3.1 source-integrity evidence for bidi controls, hidden/default-ignorable characters, Unicode tag and variation-selector payloads, and bounded mixed-script confusables. V3.2 additionally audits explicit AI attribution in source/git/media metadata and validates supported local C2PA assets with the official optional Content Authenticity Initiative library. Remote manifests and revocation endpoints are never fetched. Statistical-watermark detection remains explicitly unavailable. Licence findings are off by default (informational, not fixable); pass include_licenses: true to add them. Lists declared coverage gaps (coverage_gaps, stable ids) that make coverage partial and informational coverage_notes; accept_gaps accepts named gap ids for CI while they stay listed. Fully offline — zero network egress at scan time. Never writes to your code or repo.

codeinspectus_setupA

Inspect external-engine health and exact platform download sizes, save declined choices, or install selected Opengrep/Gitleaks/Trivy/OSV-Scanner components after explicit confirmation. Pass path (the repository to scan) to plan or install only the OSV vulnerability databases for the ecosystems its lockfiles use (and GitHub Actions advisories when it has workflows); each is listed with an estimated size (integrity: TLS + recorded SHA-256, OSV publishes no signature), and osv_ecosystems pins the exact list the user approved. Plan is offline. Install writes only to ~/.codeinspectus, verifies immutable pins/publisher provenance for engines, and never modifies the target repository.

codeinspectus_rescanA

Re-run a scan after fixes were applied and diff against a prior scan_id (or the most recent scan of the same path). Reports which findings are resolved, which remain, and which were newly introduced, plus fresh technology and native-pack execution coverage. Repository-trust artifacts are diffed separately with fail-closed resolved, remaining, introduced and not-rechecked states. Use this to verify approved fixes. Never writes to your code or repo.

codeinspectus_plan_cleanupA

Create a fresh, read-only V3.3 cleanup plan for exact verified repository-trust artifact IDs. The plan reproduces current evidence, rejects protected, ambiguous, or unsupported records, hashes every preimage, and discloses exact text edits or copy-only media metadata transformations. It never changes the repository.

codeinspectus_apply_cleanupA

Apply one exact ready V3.3 cleanup plan after explicit approval. Requires an exact artifact-ID match and confirmation, creates managed content backups, uses atomic bounded edits, creates new media copies, reruns the same repository-trust validators, and writes a content-free audit log. This modifies selected source files and/or creates named cleaned copies.

codeinspectus_rollback_cleanupA

Restore an applied V3.3 cleanup from its managed checkpoint after explicit confirmation. Rollback refuses to overwrite files or cleaned copies that changed after cleanup.

codeinspectus_compliance_reportA

Produce a per-framework code-level control-coverage view for a prior scan (NIST CSF 2.0, ISO 27001:2022, SOC 2, CIS v8.1, Essential Eight, OWASP Web/LLM). Reports 'X of N code-visible controls have findings' with the code-visible subset as the explicit denominator. This is NOT a compliance audit, certification, or attestation — code-level evidence only.

codeinspectus_explain_findingA

Return a deep explanation and full remediation plan for a single finding id from a prior scan: what the weakness is, why it matters, concrete fix steps, and references.

codeinspectus_generate_sbomA

Generate a CycloneDX or SPDX SBOM for the target project using Trivy plus the first-party offline Pub lockfile inventory, with native Pub fallback when Trivy is unavailable. Writes the SBOM file to the chosen output path and returns its location and component count. Offline.

codeinspectus_list_rulesB

List the active detectors and engine versions, the CodeInspectus detection-database version and date, Trivy vulnerability-DB freshness, bundled Pub advisory-database provenance/freshness, and the custom CodeInspectus AI-code rules and native detector packs currently shipped.

Prompts

Interactive templates invoked by user choice

NameDescription

No prompts

Resources

Contextual data attached and managed by the client

NameDescription

No resources

TDQS

A3.9/5.0

Scored across 10 tools

Disambiguation4/5

Each tool targets a distinct phase of the workflow: scanning (scan/rescan), reporting (compliance_report, explain_finding, generate_sbom), inspection (list_rules, setup), and cleanup lifecycle (plan/apply/rollback). The main overlap is between setup and list_rules, both surfacing engine/DB versions and health, which could cause occasional misselection. Otherwise boundaries are clear.

Naming Consistency4/5

All tools share the codeinspectus_ prefix and most follow a verb_noun pattern (explain_finding, generate_sbom, plan_cleanup, apply_cleanup, rollback_cleanup, compliance_report). Minor deviation: scan, setup, and rescan are bare verbs/single words rather than verb_noun, but the convention is still readable and predictable overall.

Tool Count5/5

Ten tools is well-scoped for a security scanner with scanning, reporting, engine management, and a guarded cleanup lifecycle. Each tool earns its place with no obviously redundant entries.

Completeness4/5

The surface covers the full workflow: scan, rescan/diff, finding explanation, compliance mapping, SBOM generation, engine setup, and a safe plan/apply/rollback cleanup cycle. Minor gaps exist around scan-history management (no explicit list/delete of prior scans) and configuration of scan options beyond what scan accepts, but agents can work around these via scan_id references.

Maintenance

ActivityActive
ResponsivenessLimited data