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

# Quickstart

> Connect a Product Brain workspace and complete the first useful Chain workflow.

<Warning>
  **Early access** — UI may differ by account. [**Join the Product Brain Community**](https://community.productbrain.io).
</Warning>

This is the hands-on path. If you are still trying to understand the model, start with [The Chain](/docs/the-chain).

## 0. Install the CLI (optional)

If you want terminal workflows, install **`pb`**: [Install pb](/docs/install). MCP-only users skip to step 1.

## 1. Open your workspace

Sign in at `work.productbrain.io`, open the workspace you want to connect, and use the app-provided connect command or MCP configuration.

## 2. Pick a connection path

Choose based on where your AI assistant runs:

* Use MCP when your assistant runs in Cursor, Claude, ChatGPT, or another tool that supports connected tools.
* Use the CLI when your assistant can run terminal commands in your project and you want terminal-native control.

Both paths connect to the same Product Brain workspace.

Start with one of these guides:

* [Connect Your Tools](/docs/integrations/overview)
* [Cursor](/docs/integrations/cursor)
* [Claude](/docs/integrations/claude)
* [ChatGPT](/docs/integrations/chatgpt)
* [Codex](/docs/integrations/codex)

## 3. Orient before work

Before substantive work, your agent should orient from the Chain. Orientation loads the relevant decisions, tensions, standards, and active work for the task.

What you say to your agent:

```text theme={null}
Use Product Brain to orient before we work on the onboarding flow.
```

<details>
  <summary>CLI equivalent</summary>

  ```bash theme={null}
  pb orient --task "Describe the work you are about to do"
  ```
</details>

## 4. Start a write session

When work will change code, product decisions, or reusable knowledge, start a session. In MCP-connected tools, the agent can do this for you.

What you say to your agent:

```text theme={null}
Start a session. I am working on onboarding.
```

<details>
  <summary>CLI equivalent</summary>

  ```bash theme={null}
  pb session start
  ```
</details>

## 5. Capture what changed

Capture decisions, tensions, or insights as they appear. The human says what changed; the agent handles the tool call.

What you say to your agent:

```text theme={null}
Save a decision: user-visible work ships behind feature flags until rollout confidence is high.

Save a tension: mobile auth still breaks during uncontrolled rollouts.

Save an insight: users drop at step 3 when they hit an unfinished onboarding flow.
```

<details>
  <summary>CLI equivalent</summary>

  ```bash theme={null}
  pb capture "DEC: chose X because Y"
  pb capture "TEN: blocked by Z"
  pb capture "INS: learned that A changes B"
  ```
</details>

## 6. Close the loop

Close the session so the next agent or teammate inherits current context.

What you say to your agent:

```text theme={null}
Close the Product Brain session and summarize what changed.
```

<details>
  <summary>CLI equivalent</summary>

  ```bash theme={null}
  pb session close
  ```
</details>

## FAQ

### Do I need a developer?

If your AI assistant already works inside your repo or supports MCP tools, you can usually try Product Brain yourself. Some team or company-managed AI clients may require admin permissions.

### Can I try this without changing my product?

Yes. Start by orienting and capturing one decision. You do not need to expose secrets or let an agent write anything just to understand the workflow.

### My AI client does not show Product Brain tools

Restart the AI client after changing its MCP configuration. Many clients only load tools at startup.

### The assistant says writes are blocked

Write tools require the right workspace key and a grounded session. Ask the agent to orient around the task and start a session before it writes.

### The docs show old content

Mintlify deploys from GitHub. If a docs change was committed but is not visible yet, confirm Mintlify is pointed at `docs/public`, redeploy Mintlify, or wait for the deploy to finish.