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

# Tools and permissions

> Understand tool authority, prompts, and execution boundaries

Tools let agents read files, search code, edit workspaces, run commands, fetch web content, query integrations, and perform security analysis.

## Permission outcomes

Each request is evaluated against ordered rules and resolves to:

* **Allow**: execute without prompting.
* **Ask**: pause and request your approval.
* **Deny**: reject the operation.

Rules can match tool names, filesystem patterns, commands, URLs, metadata, or agent/session policy. More specific rules should be preferred over broad permanent allowances.

## Location scope

Filesystem services, tools, providers, and permissions belong to the location that owns the session. Access outside the workspace requires explicit external-directory authority.

## Tool settlement

Tool calls are durable timeline parts with pending, running, completed, and error states. TurenOS settles in-flight calls before promoting queued input or completing a provider turn.

## Risk boundaries

TurenOS is not a VM or container sandbox. Allowed shell commands and edits run with your user privileges. Read-only analyzers do not execute the target, but their results should still be treated as evidence rather than automatic truth.

See [Trust and permissions](/security/trust-permissions) for safe defaults.