codeql-development-mcp-server

Store findings

Create annotation

List all audit findings for a repository, including any analyst notes.

Append notes to an existing audit finding. Identify the finding by findingId (preferred) or by owner+repo+sourceLocation+line.

Delete all audit findings for a repository.

Retrieve a single annotation by its numeric ID.

List annotations with optional filtering by category, entity key, or entity key prefix.

Update the content, label, or metadata of an existing annotation.

Delete annotations by ID, category, or entity key prefix.

Full-text search across annotation content, metadata, and labels using SQLite FTS (token-based MATCH; use * suffix for prefix matching, e.g.

Extract all data for a specific rule/query from multi-rule SARIF. Returns a valid SARIF JSON subset with only the matching rule definition and results.

List all rules in a SARIF file with result counts, severity, precision, and tags. Essential for discovering available rules before extraction or comparison.

Convert per-rule SARIF data to a structured markdown report with Mermaid dataflow diagrams. Renders dataflow paths as visual flowcharts.

Compare code locations of two SARIF alerts to detect overlap. Supports sink, source, any-location, full-path, and fingerprint comparison modes.

Correlate SARIF results with a git diff to classify findings as

Diff two SARIF files or cached results to find added, removed, and changed rules/results. Useful for comparing analysis across CodeQL versions, database updates, or query pack releases.

Store SARIF content in the session cache for use by other sarif_* tools. Returns a cache key that can be passed to sarifPath/cacheKey parameters of other tools.

Identify duplicate alerts across two SARIF files by comparing rules pairwise. Uses fingerprint matching first, then full-path location overlap as fallback. Useful for cleanup after query changes or pack upgrades.

Check whether cached query results exist for given parameters. Returns metadata about the cached entry without the full content.

Retrieve cached query results with optional subset selection. Supports line ranges (for graphtext/CSV) and SARIF result indices and file filtering to return only the relevant portion.

Clear cached query results by cache key, query name, database path, or clear all.

Compare cached query results across multiple databases for the same query. Useful for MRVA-style cross-repository analysis.

End a query development session with final status

Get complete details of a specific query development session

List query development sessions with optional filtering

Update the current state of a query development session

Get MCP call history for a specific session

Get test execution history for a specific session

Get quality score history for a specific session

Calculate current quality score for a session based on its state

Compare multiple query development sessions across specified dimensions

Generate aggregate insights from multiple sessions based on filters

Export session data in specified format for external analysis

Quick heuristic validation for CodeQL query structure - checks for common patterns like from/where/select clauses and metadata presence. Does NOT compile the query. For authoritative validation with actual compilation, use codeql_lsp_diagnostics instead.

Create directory structure and files for a new CodeQL query with tests

Find and track all files and directories related to a CodeQL query, including resolved metadata

List CodeQL databases discovered in configured base directories (set via CODEQL_DATABASES_BASE_DIRS env var). Returns path, language, CLI version, and creation time for each database. Use the returned database paths with codeql_query_run or codeql_database_analyze to run queries against them.

Finds startline, startcol, endline endcol of a class for quickeval

Profile the performance of a CodeQL query run against a specific database by analyzing the evaluator log JSON file

Finds startline, startcol, endline endcol of a predicate for quickeval

Read source file contents from a CodeQL database source archive (src.zip) or source directory (src/).

Search QL source files (.ql/.qll) for text or regex patterns. Returns structured results with file paths, line numbers, and optional context lines. Use this instead of grep for searching CodeQL source code.

Register a CodeQL database for use by other MCP tools.

List discovered query run result directories (set via CODEQL_QUERY_RUN_RESULTS_DIRS env var). Returns path, query name, timestamp, language, query file path, and available artifacts (evaluator-log, bqrs, sarif, query.log, summary) for each run. Filter by queryName, language, or queryPath to narrow results. Use the returned BQRS paths with codeql_bqrs_decode or codeql_bqrs_info to inspect query results.

List discovered MRVA (Multi-Repository Variant Analysis) run results (set via CODEQL_MRVA_RUN_RESULTS_DIRS env var). Returns run ID, timestamp, repositories scanned, analysis status, and available artifacts for each run.

Resolve a class or predicate symbol in a CodeQL query file for quick evaluation.

Parse CodeQL evaluator logs into a structured performance profile. Returns compact JSON with per-query summaries and top-N slowest predicates (name, duration, result size, eval order, dependency count). Full RA operations, pipeline-stage tuple progressions, and dependency lists are written to a line-indexed detail file — each predicate includes detailLines: {start, end} for targeted read_file access to its full analysis. Works with logs from codeql query run, codeql database analyze, or vscode-codeql.

Get code completions at a cursor position in a CodeQL file. Returns completion items with labels, documentation, and insert text. The file must be a .ql or .qll file. IMPORTANT: Set workspace_uri to the pack or workspace root directory for dependency resolution; without it, completions for imported libraries will be empty.

Go to the definition of a CodeQL symbol at a given position. Returns one or more file locations where the symbol is defined. Set workspace_uri to the pack root for dependency resolution.

Find all references to a CodeQL symbol at a given position. Returns file locations of all usages, including the declaration. Set workspace_uri to the pack root for dependency resolution.

List all top-level definitions (classes, predicates, modules) in a CodeQL file. Response contains location and type information unless names_only is set to true.

Authoritative syntax and semantic validation of CodeQL (QL) code via the CodeQL Language Server. Compiles the query and provides real-time diagnostics with precise error locations. Use this for accurate validation; for quick heuristic checks without compilation, use validate_codeql_query instead. Note: inline ql_code is evaluated as a virtual document and cannot resolve pack imports (e.g. `import javascript`). For validating queries with imports, use codeql_query_compile on the actual file instead.