# Use Bkper with your AI assistant

Use the AI assistant you already have to ask questions about your books, create reports and PDFs, review transactions, or build a custom workflow. You do not need to learn command-line tools or choose an AI model to get started.

> **Note: Your assistant stays your assistant**
> Your assistant continues using its own models and plan. Bkper's CLI and skill or the Bkper MCP Server give it access to Bkper; they do not replace its model with a Bkper model.
## If your assistant can run commands

Use this path when your assistant has access to files and commands in a workspace. That workspace can be on your computer or in a hosted sandbox, such as ChatGPT Work on the web. What matters is command access, not where you opened the assistant.

Paste this prompt into your command-capable assistant:

```text
Set up Bkper in this workspace using only official instructions from bkper.com.

Install the official Bkper CLI and Bkper skill, then guide me through sign-in. When `bkper auth login` shows the device code, share it with me without ending your turn or stopping the command. Keep that same process running while I authorize in my browser; once login succeeds, verify access by listing my books. Do not change any book data.
```

The assistant will:

1. Check what its workspace needs.
2. Explain any software it needs to install.
3. Ask you to complete Bkper sign-in in your browser.
4. Verify the connection with a read-only list of your books.

You are ready when the assistant can show the books available to your Bkper account without changing them. If the workspace is hosted, check whether files and sign-in persist between sessions before relying on it for ongoing work.

### Get your first useful result

Create a report:

```text
Create a profit and loss report for last month from my Bkper book. Use Bkper balances as the source of truth, explain the reporting period, and export the result as a PDF.
```

Ask about your books:

```text
Explain what changed most in my income and expenses over the last three months. Use Bkper reports for every number and state the periods you compared.
```

Review transactions:

```text
Review unchecked transactions from this month and identify anything unusual or possibly duplicated. Show me your findings without changing anything.
```

Build a custom tool:

```text
Build a simple app in this workspace that displays monthly profit and loss from my Bkper book. Add tests and let me review it before anything is deployed.
```

## Connect with the Bkper MCP Server

If your assistant cannot run commands, or you prefer quick conversational exploration, [connect to the Bkper MCP Server](https://bkper.com/apps/bkper-mcp.md) without installing the CLI. After authorizing your Bkper account, use the same report, question, and review prompts above. This path works in compatible web or desktop assistants; where you opened the assistant does not determine which Bkper connection to use.

The MCP Server is a good fit for:

- asking questions about Book structure;
- listing Accounts, Groups, Transactions, and balances;
- creating reports from Bkper balance data;
- reviewing suspicious or unchecked Transactions;
- exploring your data before deciding what to do next.

The assistant may be able to create a downloadable file using its own file tools. The financial values must still come from Bkper records and balance reports. The Bkper MCP Server provides tools; your assistant continues using its own models.

## Prefer a dedicated Bkper workspace?

Bkper CLI also includes [Bkper CLI Agent](https://bkper.com/docs/ai/bkper-cli-agent.md), an optional terminal workspace with Bkper context, Book access, local files, commands, scripts, and tests. Run `bkper` to open it.

Bkper CLI Agent comes with selected [Bkper AI models and usage allowance](https://bkper.com/docs/ai/models.md) already configured. You can also connect an external model provider. Otherwise, keep using ChatGPT or Claude—there is no need to switch.

## Keep financial work reviewable

- Ask the assistant to use Bkper as the source of truth for balances and reports.
- Do not accept model arithmetic as the final source for statements, reconciliations, or taxes.
- Keep AI-derived Transactions as drafts until a person reviews and posts them.
- Require an explanation and explicit confirmation before any Book change.
- Review generated reports, files, tests, and app changes before relying on them.
- Treat app deployment as a separate action that requires your approval.

## What each part does

| Path                                            | Assistant                             | Bkper access              | Models                                                        |
| ----------------------------------------------- | ------------------------------------- | ------------------------- | ------------------------------------------------------------- |
| **Command-capable assistant (local or hosted)** | ChatGPT, Claude, or another assistant | Bkper CLI and skill       | Provided by your assistant                                    |
| **MCP-capable assistant (web or desktop)**      | ChatGPT, Claude, or another assistant | Bkper MCP Server          | Provided by your assistant                                    |
| **Dedicated terminal workspace**                | Bkper CLI Agent                       | Built-in Bkper CLI access | Included Bkper AI models, or an external provider you connect |

## Technical connection details

You do not need these implementation details to get started.

### CLI and skill connection

This path uses the official Bkper CLI and Bkper skill, whether the workspace is on your computer or in a hosted sandbox. Your assistant can combine Bkper data with workspace files, scripts, tests, and generated artifacts. Command-line tools can process intermediate data without sending every step back into the conversation.

### Bkper MCP Server

The Bkper MCP Server exposes typed Bkper tools through your existing permissions. Each tool result returns to the conversation, making this path well suited to interactive exploration.

## Next steps

- [Install or reference the official Bkper tools](https://bkper.com/apps/bkper-cli.md)
- [Connect through the Bkper MCP Server](https://bkper.com/apps/bkper-mcp.md)
- [Review CLI and sandbox security](https://bkper.com/docs/ai/cli-agent-security.md)
- [Build with a coding agent](https://bkper.com/docs/ai/coding-agents.md)