An MCP client (i.e. calling LLM) should be able to provide additional arguments such as --sarif-include-query-help=always and --no-sarif-minify while having the codeql_database_analyze tool run without error. The truly desired behavior is that any argument supported by the codeql database analyze CLI command should be supportable via the codeql_database_analyze MCP tool.
$ codeql database analyze -h -vv
Usage: codeql database analyze [OPTIONS] -- <database> [<query|dir|suite|pack>...]
Analyze a database, producing meaningful results in the context of the source code.
Run a query suite (or some individual queries) against a CodeQL database, producing results, styled as alerts or paths, in SARIF or another interpreted format.
This command combines the effect of the codeql database run-queries and codeql database interpret-results commands. If you want to run queries whose results don't meet the requirements for being interpreted as source-code alerts, use codeql database run-queries or codeql query run instead, and then
codeql bqrs decode to convert the raw results to a readable notation.
<database> [Mandatory] Path to the CodeQL database to query.
[<query|dir|suite|pack>...]
Queries to execute. Each argument is in the form scope/name@range:path where:
* scope/name is the qualified name of a CodeQL pack.
* range is a semver range.
* path is a file system path.
If a scope/name is specified, the range and path are optional. A missing range implies the latest version of the specified pack. A missing path implies the default query suite of the specified pack.
The path can be one of a *.ql query file, a directory containing one or more queries, or a .qls query suite file. If there is no pack name specified, then a path must be provided, and will be interpreted relative to the current working directory of the current process.
To specify a path that contains a literal @ or :, use path: as a prefix to the argument, like this: path:directory/with:and@/chars.
If a scope/name and path are specified, then the path cannot be absolute. It is considered relative to the root of the CodeQL pack.
If no queries are specified, the CLI will automatically determine a suitable set of queries to run. In particular, if a Code Scanning configuration file was specified at database creation time using --codescanning-config then the queries from this will be used. Otherwise,
the default queries for the language being analyzed will be used.
--format=<format> [Mandatory] The format in which to write the results. One of:
csv: Formatted comma-separated values, including columns with both rule and alert metadata.
sarif-latest: Static Analysis Results Interchange Format (SARIF), a JSON-based format for describing static analysis results. This format option uses the most recent supported version (v2.1.0). This option is not suitable for use in automation as it will produce different
versions of SARIF between different CodeQL versions.
sarifv2.1.0: SARIF v2.1.0.
graphtext: A textual format representing a graph. Only compatible with queries with @kind graph.
dgml: Directed Graph Markup Language, an XML-based format for describing graphs. Only compatible with queries with @kind graph.
dot: Graphviz DOT language, a text-based format for describing graphs. Only compatible with queries with @kind graph.
-o, --output=<output> [Mandatory] The output path to write results to. For graph formats this should be a directory, and the result (or results if this command supports interpreting more than one query) will be written within that directory.
--[no-]rerun Evaluate even queries that seem to have a BQRS result stored in the database already.
--no-print-diagnostics-summary
Don't print a summary of the analyzed diagnostics to standard output.
--no-print-metrics-summary
Don't print a summary of the analyzed metrics to standard output.
--max-paths=<maxPaths> The maximum number of paths to produce for each alert with paths. (Default: 4)
--[no-]sarif-add-file-contents
[SARIF formats only] Include the full file contents for all files referenced in at least one result.
--[no-]sarif-add-snippets
[SARIF formats only] Include code snippets for each location mentioned in the results, with two lines of context before and after the reported location.
--[no-]sarif-add-query-help
[SARIF formats only] [Deprecated] Include Markdown query help for all queries. It loads query help for /path/to/query.ql from the /path/to/query.md file. If this flag is not supplied the default behavior is to include help only for custom queries i.e. those in query packs
which are not of the form `codeql/<lang&rt;-queries`. This option has no effect when passed to codeql bqrs interpret.
--sarif-include-query-help=<mode>
[SARIF formats only] Specify whether to include query help in the SARIF output. One of:
always: Include query help for all queries.
custom_queries_only (default): Include query help only for custom queries i.e. those in query packs which are not of the form `codeql/<lang&rt;-queries`.
never: Do not include query help for any queries.
This option has no effect when passed to codeql bqrs interpret.
--no-sarif-include-alert-provenance
[Advanced] [SARIF formats only] Do not include alert provenance information in the SARIF output.
--[no-]sarif-group-rules-by-pack
[SARIF formats only] Place the rule object for each query under its corresponding QL pack in the <run>.tool.extensions property. This option has no effect when passed to codeql bqrs interpret.
--[no-]sarif-multicause-markdown
[SARIF formats only] For alerts that have multiple causes, include them as a Markdown-formatted itemized list in the output in addition to as a plain string.
--no-sarif-minify [SARIF formats only] Produce pretty-printed SARIF output. By default, SARIF output is minified to reduce the size of the output file.
--sarif-run-property=<String=String>
[SARIF formats only] A key value pair to add to the generated SARIF 'run' property bag. Can be repeated.
--no-group-results [SARIF formats only] Produce one result per message, rather than one result per unique location.
--csv-location-format=<csvLocationFormat>
The format in which to produce locations in CSV output. One of: uri, line-column, offset-length. (Default: line-column)
--dot-location-url-format=<dotLocationUrlFormat>
A format string defining the format in which to produce file location URLs in DOT output. The following place holders can be used {path} {start:line} {start:column} {end:line} {end:column}, {offset}, {length}
--no-sarif-add-baseline-file-info
[Wizards only!] [SARIF formats only] Do not include baseline file information in the SARIF output.
--[no-]sublanguage-file-coverage
[GitHub.com and GitHub Enterprise Server v3.12.0+ only] Use sub-language file coverage information. This calculates, displays, and exports separate file coverage information for languages which share a CodeQL extractor like C and C++, Java and Kotlin, and JavaScript and
TypeScript.
--no-sarif-include-diagnostics
[Wizards only!] [SARIF formats only] Exclude diagnostic information from the SARIF output.
--sarif-category=<category>
[SARIF formats only] [Recommended] Specify a category for this analysis to include in the SARIF output. A category can be used to distinguish multiple analyses performed on the same commit and repository, but on different languages or different parts of the code.
If you analyze the same version of a code base in several different ways (e.g., for different languages) and upload the results to GitHub for presentation in Code Scanning, this value should differ between each of the analyses, which tells Code Scanning that the analyses
supplement rather than supersede each other. (The values should be consistent between runs of the same analysis for different versions of the code base.)
This value will appear (with a trailing slash appended if not already present) as the <run>.automationDetails.id property.
--sarif-codescanning-config=<userConfig>
[Wizards only!] [SARIF formats only] Code Scanning configuration file specifying how the CodeQL database was created and what queries were run.
--no-database-extension-packs
[Advanced] Omit extension packs stored in the database during database creation, either from a Code Scanning configuration file or from extension files stored in the 'extensions' directory of the analyzed codebase.
--no-database-threat-models
[Advanced] Omit threat model configuration stored in the database during database creation from a Code Scanning configuration file.
--[no-]download Download any missing queries before analyzing.
Options to control the model packs to be used:
--model-packs=<name@range>...
A list of CodeQL pack names, each with an optional version range, to be used as model packs to customize the queries that are about to be evaluated.
Options to control the threat models to be used:
--threat-model=<name>...
A list of threat models to enable or disable.
The argument is the name of a threat model, optionally preceded by a '!'. If no '!' is present, the named threat model and all of its descendants are enabled. If a '!' is present, the named threat model and all of its descendants are disabled.
The 'default' threat model is enabled by default, but can be disabled by specifying '--threat-model !default'.
The 'all' threat model can be used to enable or disable all threat models.
The --threat-model options are processed in order. For example, '--threat-model local --threat-model !environment' enables all of the threat models in the 'local' group except for the 'environment' threat model.
This option only has an effect for languages that support threat models.
Options to control the query evaluator:
--[no-]debug [Wizards only!] Include additional debugging data in log output.
--[no-]tuple-counting [Advanced] Display tuple counts for each evaluation step in the query evaluator logs. If the --evaluator-log option is provided, tuple counts will be included in both the text-based and structured JSON logs produced by the command. (This can be useful for performance
optimization of complex QL code).
--timeout=<seconds> [Advanced] Set the timeout length for query evaluation, in seconds.
The timeout feature is intended to catch cases where a complex query would take "forever" to evaluate. It is not an effective way to limit the total amount of time the query evaluation can take. The evaluation will be allowed to continue as long as each separately timed
part of the computation completes within the timeout. Currently these separately timed parts are "RA layers" of the optimized query, but that might change in the future.
If no timeout is specified, or is given as 0, no timeout will be set (except for codeql test run, where the default timeout is 5 minutes).
-j, --threads=<num> Use this many threads to evaluate queries.
Defaults to 1. You can pass 0 to use one thread per core on the machine, or -N to leave N cores unused (except still use at least one thread).
--no-intra-layer-parallelism
[Wizards only!] Controls whether the scheduler should recruit additional worker threads to evaluate individual layers in parallel. For internal testing purposes only.
--[no-]evaluate-as-overlay
[Wizards only!] Override the automatic detection of whether to evaluate queries in 'overlay' mode or not.
This is an experimental feature still under development; do not use.
--[no-]cache-at-frontier
[Wizards only!] Force evaluation and caching of intermediate results that will be useful for future overlay evaluation. In codeql database analyze and codeql database run-queries, this is turned on by default when evaluating against an overlay-prepared database that
doesn't yet have an overlay, so it does not usually need to be specified explicitly. This is an experimental feature still under development; do not use.
--[no-]warm-cache-only [Wizards only!] When enabled, the evaluator will still ensure that explicitly-cached predicates are evaluated and written to the cache, but no query results will be produced.
--[no-]force-dbscheme [Wizards only!] Force execution of queries that don't appear to be compatible with the dataset. May lead to bad results or execution failures.
--no-use-precompiled [Wizards only!] Disable the use of any precompiled *.qlx files found next to *.ql files being evaluated. (Note that this option is intended as an escape hatch for extraordinary situations, but in most situations the appropriate way to avoid using *.qlx files would be to
simply delete them on disk).
--[no-]save-cache [Deprecated] [Advanced] This flag does nothing.
--[no-]expect-discarded-cache
[Advanced] Make decisions about which predicates to evaluate, and what to write to the disk cache, based on the assumption that the cache will be discarded after the queries have been executed.
--[no-]keep-full-cache [Advanced] Don't clean up the disk cache after evaluation completes. This may save time if you're going to do codeql dataset cleanup or codeql database cleanup afterwards anyway.
--[no-]crash-below-frontier
[Wizards only!] Crash with a graph dump if we try to evaluate RA below the overlay frontier.
For internal debugging only: do not use.
--[no-]deterministic-scheduling
[Wizards only!] Disable optimizations in the scheduler that can make it timing-dependent whether certain parts of the evaluation are recognized as superfluous before they start. This is only really useful to do for internal performace testing.
--max-disk-cache=<MB> Set the maximum amount of space that the disk cache for intermediate query results can use.
If this size is not configured explicitly, the evaluator will try to use a "reasonable" amount of cache space, based on the size of the dataset and the complexity of the queries. Explicitly setting a higher limit than this default usage will enable additional caching which
can speed up later queries.
--min-disk-free=<MB> [Advanced] Set target amount of free space on file system.
If --max-disk-cache is not given, the evaluator will try hard to curtail disk cache usage if the free space on the file system drops below this value.
--min-disk-free-pct=<pct>
[Advanced] Set target fraction of free space on file system.
If --max-disk-cache is not given, the evaluator will try hard to curtail disk cache usage if the free space on the file system drops below this percentage.
--external=<pred>=<file.csv>
A CSV file that contains rows for external predicate <pred>. Multiple --external options can be supplied.
--xterm-progress=<mode>
[Advanced] Controls whether to show progress tracking during QL evaluation using xterm control sequences. Possible values are:
no: Never produce fancy progress; assume a dumb terminal.
auto (default): Autodetect whether the command is running in an appropriate terminal.
yes: Assume the terminal can understand xterm control sequences. The feature still depends on being able to autodetect the size of the terminal (which is not implemented on Windows, sorry), and will also be disabled if -q is given.
25x80 (or similar): Like yes, and also explicitly give the size of the terminal. (Unlike yes, this should work on Windows.)
25x80:/dev/pts/17 (or similar): show fancy progress on a different terminal than stderr. Mostly useful for internal testing.
Options for controlling outputting of structured evaluator logs:
--evaluator-log=<file> [Advanced] Output structured logs about evaluator performance to the given file. The format of this log file is subject to change with no notice, but will be a stream of JSON objects separated by either two newline characters (by default) or one if the
--evaluator-log-minify option is passed. Please use codeql generate log-summary <file> to produce a more stable summary of this file, and avoid parsing the file directly. The file will be overwritten if it already exists.
--evaluator-log-minify [Advanced] If the --evaluator-log option is passed, also passing this option will minimize the size of the JSON log produced, at the expense of making it much less human readable.
--evaluator-log-level=<n>
[Wizards only!] If the --evaluator-log option is passed, also passing this option will configure the verbosity of the JSON log produced. This should be an integer between 1 and 5, with higher values representing a higher verbosity. Currently the following additional values
will be included at each level (in addition to all information already present in preceding levels):
* Level 1: Include basic information about each computation performed by the evaluator.
* Level 2: Include the dependencies of every layer that is computed.
* Level 3: Include details of where each layer is used.
* Level 4: Record an event when a cache lookup misses.
* Level 5 (default): Output the full RA of any relations that are evaluated.
Options to control RAM usage:
-M, --ram=<MB> The query evaluator will try hard to keep its total memory footprint below this value. (However, for large databases it is possible that the threshold may be broken by file-backed memory maps, which can be swapped to disk in case of memory pressure).
The value should be at least 2048 MB; smaller values will be transparently rounded up.
--heap-ram=<MB> [Wizards only!] Explicitly specify the amount of RAM from the Java heap the evaluator cache will use. This will be taken from the total --ram amount. Be sure to leave some RAM for non-cache purposes. Beware that it does not make much sense to specify this without also
giving --ram explicitly.
--off-heap-ram=<MB> [Wizards only!] Set amount of off-heap RAM. This will be taken from the total --ram amount.
Options to control QL compilation:
--allow-experimental=<feature>[,<feature>...]
[Wizards only!] Allow the use of an experimental QL language feature whose compatibility with earlier or later releases of the CodeQL CLI cannot be reliably expected. The selection of features governed by this option is not documented, as the option is only expected to be
used by the internal teams developing the features.
This option is required when queries that use the feature are being published in a query pack; in all other cases, omitting the option merely produces a compiler warning.
--dynamic-join-order-mode=<dynamicJoinMode>
[Wizards only!] Enable the dynamic join ordering experimental feature. This feature is experimental and a work in progress and may be removed or changed in future releases.
--warnings=<mode> How to handle warnings from the QL compiler. One of:
hide: Suppress warnings.
show (default): Print warnings but continue with compilation.
error: Treat warnings as errors.
--no-debug-info Don't emit source location info in RA for debugging.
--no-dil-debug-types [Wizards only!] Don't emit debug types in pretty-printed DIL.
--[no-]fast-compilation
[Deprecated] [Advanced] Omit particularly slow optimization steps.
--no-release-compatibility
[Advanced] Use the newest compiler features, at the cost of portability.
From time to time, new QL language features and evaluator optimizations will be supported by the QL evaluator a few releases before they are enabled by default in the QL compiler. This helps ensure that the performance you experience when developing queries in the newest
CodeQL release can be matched by slightly older releases that may still be in use for Code Scanning or CI integrations.
If you do not care about your queries being compatible with other (earlier or later) CodeQL releases, you can sometimes achieve a small amount of extra performance by using this flag to enable recent improvements in the compiler early.
In releases where there are no recent improvements to enable, this option silently does nothing. Thus it is safe to set it once and for all in your global CodeQL config file.
--[no-]local-checking Only perform initial checks on the part of the QL source that is used.
--no-metadata-verification
Don't check embedded query metadata in QLDoc comments for validity.
--no-to-string [Wizards only!] Don't generate code to render all result columns as strings.
--no-get-url [Wizards only!] Don't generate code to render URL for result columns with locations.
--no-fake-url [Wizards only!] Don't generate code to give a fake URL for results that do not have them.
--no-default-to-string [Wizards only!] Don't generate code to give a default toString for results that do not have them.
--no-parallel-compilation
[Wizards only!] Disable parallel compilation. (Should never be necessary, but who knows? It's a pretty recent feature).
--compilation-cache-size=<MB>
[Advanced] Override the default maximum size for a compilation cache directory.
--fail-on-ambiguous-relation-name
[Advanced] Fail compilation if an ambiguous relation name is generated during compilation.
Options to set up compilation environment:
--search-path=<dir>[:<dir>...]
A list of directories under which QL packs may be found. Each directory can either be a QL pack (or bundle of packs containing a .codeqlmanifest.json file at the root) or the immediate parent of one or more such directories.
If the path contains more than one directory, their order defines precedence between them: when a pack name that must be resolved is matched in more than one of the directory trees, the one given first wins.
Pointing this at a checkout of the open-source CodeQL repository ought to work when querying one of the languages that live there.
If you have checked out the CodeQL repository as a sibling of the unpacked CodeQL toolchain, you don't need to give this option; such sibling directories will always be searched for QL packs that cannot be found otherwise. (If this default does not work, it is strongly
recommended to set up --search-path once and for all in a per-user configuration file).
--additional-packs=<dir>[:<dir>...]
If this list of directories is given, they will be searched for packs before the ones in --search-path. The order between these doesn't matter; it is an error if a pack name is found in two different places through this list.
This is useful if you're temporarily developing a new version of a pack that also appears in the default path. On the other hand, it is not recommended to override this option in a config file; some internal actions will add this option on the fly, overriding any
configured value.
--library-path-dependency=<qlpack>
[Wizards only!] Pretend that the named QL pack is given in the libraryPathDependencies property of the QL pack that contains the query.
This option exists only to support a few internal use cases. It is highly improbable that you actually need it. Please declare your dependencies in the QL pack itself instead.
--library-path=<dir>[:<dir>...]
[Advanced] An optional list of directories that will be added to the raw import search path for QL libraries. This should only be used if you're using QL libraries that have not been packaged as QL packs.
--full-library-path=<dir>[:<dir>...]
[Wizards only!] A list of directories that give the full path for QL libraries that will be used to compile a query. This overrides the entire default path computation.
This option is internal and unstable; it may stop working the same way in new versions without much warning.
--dbscheme=<file> [Advanced] Explicitly define which dbscheme queries should be compiled against. This should only be given by callers that are extremely sure what they're doing.
--compilation-cache=<dir>
[Advanced] Specify an additional directory to use as a compilation cache.
--no-default-compilation-cache
[Advanced] Don't use compilation caches in standard locations such as in the QL pack containing the query or in the CodeQL toolchain directory.
Options for configuring the CodeQL package manager:
--qlconfig-file=<qlconfig-file>
[Wizards only!] The path to the qlconfig.yml file containing configuration.
--no-use-global-qlconfig
[Wizards only!] Do not read the shared global configuration file, and use only local configuration files.
--registries-auth-stdin
Authenticate to GitHub Enterprise Server Container registries by passing a comma-separated list of <registry_url>=<token> pairs.
For example, you can pass "https://containers.GHEHOSTNAME1/v2/=TOKEN1,https://containers.GHEHOSTNAME2/v2/=TOKEN2" to authenticate to two GitHub Enterprise Server instances.
This overrides the CODEQL_REGISTRIES_AUTH and GITHUB_TOKEN environment variables. If you only need to authenticate to the github.com Container registry, you can instead authenticate using the simpler --github-auth-stdin option.
--github-auth-stdin Authenticate to the github.com Container registry by passing a github.com GitHub Apps token or personal access token via standard input.
To authenticate to GitHub Enterprise Server Container registries, pass --registries-auth-stdin or use the CODEQL_REGISTRIES_AUTH environment variable.
This overrides the GITHUB_TOKEN environment variable.
Common options:
-h, --help Show this help text.
-J=<opt> [Advanced] Give option to the JVM running the command.
(Beware that options containing spaces will not be handled correctly.)
-v, --verbose Incrementally increase the number of progress messages printed.
-q, --quiet Incrementally decrease the number of progress messages printed.
--verbosity=<level> [Advanced] Explicitly set the verbosity level to one of errors, warnings, progress, progress+, progress++, progress+++. Overrides -v and -q.
--logdir=<dir> [Advanced] Write detailed logs to one or more files in the given directory, with generated names that include timestamps and the name of the running subcommand.
(To write a log file with a name you have full control over, instead give --log-to-stderr and redirect stderr as desired.)
--[no-]log-to-stderr [Wizards only!] Print detailed logs to standard error.
--loglevel=<level> [Wizards only!] Set the logging level of the detailed logs to one of OFF, ERROR, WARN, INFO, DEBUG, TRACE, or ALL.
--common-caches=<dir> [Advanced] Controls the location of cached data on disk that will persist between several runs of the CLI, such as downloaded QL packs and compiled query plans. If not set explicitly, this defaults to a directory named .codeql in the user's home directory; it will be
created if it doesn't already exist.
Primitive Type
Tool (Interactive function)
Primitive Name
codeql_database_analyze
Update Type
Bug Fix
Current Behavior
The additional arguments support for the
codeql_database_analyzetool is insufficient, where a client cannot provide an additional argument with a value such as--sarif-include-query-help=always.Desired Behavior
An MCP client (i.e. calling LLM) should be able to provide additional arguments such as
--sarif-include-query-help=alwaysand--no-sarif-minifywhile having thecodeql_database_analyzetool run without error. The truly desired behavior is that any argument supported by thecodeql database analyzeCLI command should be supportable via thecodeql_database_analyzeMCP tool.Example Inputs
No response
Expected Outputs
No response
Test Scenarios
No response
Documentation or Reference Materials
From
codeql database analyze -h -vv:Use Case Improvement
No response
Bug Reproduction Steps (If applicable)
No response
Error Messages or Logs (If applicable)
No response
References
No response
Priority Level
Medium (Notable improvement)
Breaking Changes Acceptable
No - Must maintain full backward compatibility
Update Checklist (Auto-populated for LLM guidance)
Code of Conduct