AI Glot websiteOpen AI Glot
Operate AI GlotMCP safety and permissions
Operate AI Glot

MCP safety and permissions

Give AI agents minimum AI Glot access, understand read and destructive annotations, and bound the effect of untrusted workspace content.

An MCP client acts with the scopes you approve. Start read-only, add write access only for a concrete use case, and review destructive calls.

Use caseScopes
Reportingaccount:read, usage:read, batches:read
Download assistantaccount:read, batches:read
Translation housekeepingAdd batches:write
Terminology assistantglossaries:read, optionally glossaries:write

The workspace member ceiling still applies to OAuth. An administrator can keep a person’s connections read-only regardless of what the client requests.

Tool annotations

Standard tools declare whether they are read-only, idempotent or destructive. Renaming and archiving are non-destructive housekeeping; archive can be reversed. delete_glossary is destructive and cannot be undone.

Code Mode’s execute can reach any operation permitted by the connection, including deletion, so it carries the most cautious annotation even when a particular script only reads.

Prompt injection boundary

Batch names and glossary terms are workspace content written by people. An agent may read that content and could be influenced by malicious instructions embedded in it. The sandbox protects the infrastructure from model-written code, but it cannot make untrusted text stop being text.

The practical boundary is permission: an influenced agent can only call the same workspace operations already granted to the connection, and every call is rate limited. Do not grant glossary write access to an agent that only needs reporting.

Before approving a connection

  • Confirm the URL is exactly https://mcp.ai-glot.com/mcp or /mcp/code.
  • Review the workspace name and requested scopes on the consent screen.
  • Prefer OAuth for a person and a dedicated API key for a service.
  • Revoke unused connections from Settings → API.
  • Review write and delete confirmations in the client rather than enabling blanket approval.

Use these docs with your AI tools

An AI agent can read this documentation directly. You do not need an account or an API key. Everything here is public and read-only.

Query these docs via MCP

Recommended

Add this server to Claude, Claude Code, Cursor, Mistral, or any tool that supports MCP. Your agent can then search AI Glot Docs documentation and read it in full, instead of answering from memory.

https://ai-glot.com/docs/mcp
  • searchFind the passages that answer a question.
  • fetchRead one page in full, as Markdown.
  • list_pagesSee every page in this documentation.

Query these docs over HTTP

The same tools also work as plain web requests. Use this for scripts, or for any tool that does not support MCP. There is one endpoint per tool. Arguments go in the query string, and the answer comes back as JSON.

https://ai-glot.com/docs/api/docs/search?query=custom+domain

Read the OpenAPI description. It is built from the same definitions as the tools, so it always matches what the endpoints do.

Read these docs as Markdown

Add .md to any page URL to get its Markdown source. You can also send the headerAccept: text/markdown to the page URL itself.

To read the whole documentation in one file, open llms-full.txt. For a short index of every page, open llms.txt.