# Truto Assistant

> Source: https://truto.one/docs/guides/assistant/truto-assistant/

Truto Assistant is an AI-powered chat sidebar built into the Truto dashboard. It lets you explore your workspace, run platform operations, and get answers about integrations and connected accounts using natural language.

## Features

- **Natural language queries** — ask about your integrations, connected accounts, sync jobs, unified APIs, and more without memorizing API paths or navigation.
- **Platform operations** — list accounts, check sync job status, query data, and inspect integration configuration directly from chat.
- **Deep links** — responses include links to relevant pages in the dashboard so you can jump straight to what you need.
- **Tool approval** — write and destructive operations (create, update, delete) require explicit approval before executing.
- **Workspace-aware** — the assistant knows about your team's integrations, connected accounts, and configuration.

## Setup

1. Navigate to **Settings → Assistant** (the **Truto Assistant** page).
2. Toggle **Enable Truto Assistant** on.
3. Enter your **Anthropic API key** — Truto uses BYOK (Bring Your Own Key), so usage is billed directly to your Anthropic account.
4. Click **Test Key** to verify the key works.
5. Choose a default model (Claude Sonnet 4.6 is selected by default).

Only team admins can configure assistant settings.

## Usage

1. Open **Truto Assistant** from the dashboard sidebar to open the chat panel.
2. Type a natural language question or command — for example, "List my connected HubSpot accounts" or "What sync jobs are failing?"
3. The assistant will call platform tools to gather information. Read operations execute immediately; write operations show an approval prompt first.
4. Use the model selector in the chat header to switch between models mid-conversation.
5. The context meter in the header shows how much of the model's context window is in use.

## Available models

| Model | Strengths |
| --- | --- |
| Claude Sonnet 4.6 (default) | Balanced speed and capability — recommended for most tasks |
| Claude Opus 4.6 | Most capable reasoning — best for complex multi-step operations |
| Claude Haiku 4.5 | Fast and economical — good for simple lookups |

You can switch models at any time using the selector in the chat header.

## Security

- **BYOK** — your Anthropic API key is stored encrypted and never shared with other teams or used outside your workspace.
- **Write approval** — any tool call that would create, update, or delete a resource requires you to click Approve before it executes.
- **Workspace scope** — the assistant can only access data within your team's workspace. It cannot see other teams' resources.
- **No training** — conversations are sent to Anthropic's API under their standard data policy (messages are not used to train models).

## Related

- [Platform MCP Server](/docs/guides/assistant/platform-mcp-server) — connect external AI tools (Cursor, Claude Desktop) to your workspace via MCP.