---
title: Titan overview
description: What Titan governs on Titan Cloud, what it does not control, and
  how to reach its administration API.
url: https://hexelstudio.com/docs/titan/overview
---

Titan decides whether each action an AI agent takes through its tools may run, before it runs. Every governed action is checked against your organization's policy, can wait for a person's approval, can be stopped at once, and is recorded in a tamper-evident log, the Agent Ledger.

Titan is in private preview. These pages describe what Titan Cloud does today.

## What Titan governs

| Feature | What it does | Page |
| --- | --- | --- |
| Approvals | A policy rule makes an action wait until enough approvers decide it in the console, the CLI or the API | [Approvals](https://hexelstudio.com/docs/titan/approvals) |
| Environment freeze | Refuse every change in an environment while it is frozen; reads still run | [Environment freeze](https://hexelstudio.com/docs/titan/environment-freeze) |
| Kill switch and session controls | Stop an agent, a session or a wider scope at once; suspend and resume a session | [Kill switch](https://hexelstudio.com/docs/titan/kill-switch) |
| Coding agents | Govern Claude Code, Codex and Kiro on developer machines with managed hooks and signed local tool rules | [Coding agents](https://hexelstudio.com/docs/titan/coding-agents) |
| Agent Ledger | A hash-chained record of every decision, execution and control, which you can read and verify | [Agent Ledger](https://hexelstudio.com/docs/titan/agent-ledger) |

## What Titan does not control

Titan governs what agents do **through their tools**: calls to Titan-governed capabilities, and on developer machines the built-in tool calls that pass through the coding agent's hooks. It is not endpoint protection.

- It does not control a process that does not go through Titan: a script run by hand, a tool run outside the agent, or a coding agent started without its Titan hooks (for example by running the real binary by its full path).
- It does not inspect or block network traffic, files or credentials that an agent can reach directly on its machine.
- Titan's protection for those paths is that governed tools hold no standing credentials on the device: an agent that goes around Titan has nothing to act with on the systems Titan governs.

Use Titan together with your endpoint security, not instead of it.

## The administration API

Administrators use the console (**Titan** in [console.hexelstudio.com](https://console.hexelstudio.com)), the [Hexel CLI](https://hexelstudio.com/docs/cli/overview) (`hexel titan ...`) or the API.

- **Base URL:** `https://api.hexelstudio.com/v1/titan/`. Approvals, sessions, kill and the Ledger are under `/v1/titan/admin/`; deployment settings such as environment states are under `/v1/titan/management/`.
- **Authentication:** an access token for a member or a service user of your organization (see [Authenticate](https://hexelstudio.com/docs/identity/authenticate)). Each operation needs a permission, named on its page, for example `titan:approval:decide`.
- **`Titan-Enforcement-Deployment` header:** every `/v1/titan/admin/` call names the Titan deployment it acts on, by its `td-...` enforcement id. Titan Cloud serves many organizations from one plane, so there is no default; a missing or unknown id is refused.

```bash
curl https://api.hexelstudio.com/v1/titan/admin/approvals?state=pending \
  -H "Authorization: Bearer $ACCESS_TOKEN" \
  -H "Titan-Enforcement-Deployment: $TITAN_ENFORCEMENT_ID"
```

## Not available on Titan Cloud yet

Some Titan features are built but not turned on for customers on Titan Cloud, so they are not described here. Among them are durable sessions (sessions that outlive one process and can be resumed elsewhere), approvals decided by the person an agent acts for on their own phone, and approval requests delivered into Slack. Contact Hexel Studio if you need one of them.