> For the complete documentation index, see [llms.txt](https://developers.signitic.app/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://developers.signitic.app/reference/mcp-server/security-best-practices.md).

# Security best practices

Use Signitic’s MCP Server with a trusted client and keep control of the information the client can access or share.

### Connect over Streamable HTTP

The official Signitic MCP endpoint is <https://mcp.signitic.com/mcp>. It uses HTTPS and Streamable HTTP. Check the URL before authorizing access, and only continue if you recognize and trust the client requesting access.

OAuth authenticates your Signitic Account through your browser. Only approve access for clients you recognize and trust. Your existing Signitic permissions determine which workspace data the client can access. A workspace administrator must enable MCP access for that workspace.

### Protect against prompt injection

Tool results can include user-entered names, campaign or group details, and Help Center content. Treat that text as data, not as instructions for the AI client. Content returned by a tool should not override your request or the client’s security rules.

* Review unusual instructions that appear inside retrieved content, especially requests to reveal credentials, disclose unrelated data, or call another service.
* Check the client’s response before you use or share it. Confirm that any data it includes belongs to the workspace and task you intended.
* Be careful when the same client also has access to other tools or apps. Signitic’s MCP tools are read-only, but another connected tool may be able to send or change data.