Report vulnerabilities privately through GitHub's private vulnerability reporting. Do not open a public issue, and do not include a working API key in the report.

You can expect an acknowledgement within seven days and an assessment within thirty. If a fix is warranted, it ships in a release with the advisory published alongside it.

• Code in this repository, including the Node.js wrapper and the Python server.
• The published npm package @trsdn/mistraldocai-mcp-server.
• The release and CI workflows in .github/workflows/.

• Vulnerabilities in Mistral AI's service. Report those to Mistral AI.
• Vulnerabilities in upstream dependencies, unless this project's use of them makes an otherwise unexploitable issue exploitable. Report those upstream.
• The consequences of a user configuring a key with more privilege than the server needs.

The server reads MISTRAL_API_KEY from the environment or from ~/.mistraldocai-mcp/.env. That file is outside the repository and outside the installed package, and it is created with your user's default permissions.

The key is sent only to Mistral AI's API. It is never logged: error paths report the failure without the credential. If you believe a key has been exposed, revoke it in the Mistral console before reporting.

The other credential this project holds is the npm publishing identity used by release.yml: OIDC trusted publishing, which mints a short-lived token per run and leaves nothing standing to rotate. The repository also carries an NPM_TOKEN secret from before trusted publishing was adopted; it is unused by the current workflow and should be deleted by the maintainer. If it is ever found to be exposed, revoke it on npmjs.com under access tokens.

Note that documents you process are transmitted to Mistral AI. See Data and privacy before processing anything confidential.