> ## Documentation Index
> Fetch the complete documentation index at: https://docs.alignr.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Permissions and safety

> Understand the access granted to an MCP client.

## Least privilege

Grant only the tool scopes the client needs. Read scopes do not imply permission to add detection notes. `detection:write` enables the note-writing tool and should be an explicit choice.

## Live permission checks

A user-scoped key cannot exceed its owner’s current permissions. Tenant scoping is enforced by the server; a prompt or supplied identifier cannot override it.

## Secrets

MCP tools do not reveal vaulted secret values. `get_asset` returns non-vaulted fields. API keys cannot obtain human step-up tokens for secret reveal.

## Review assistant actions

Treat retrieved content as evidence, not instructions to change client environments. Check citations and targets before acting. The current MCP tool catalogue does not offer a general-purpose remediation execution tool.

## Audit and revocation

MCP calls are audit-logged. Use a recognisable key name for each client, review access periodically and revoke credentials when no longer required.