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

# Why Braintrust?

> How Braintrust compares to Claude Code, OpenClaw, OpenHands, and other AI agent frameworks.

<div className="not-prose relative overflow-hidden rounded-xl border border-[#007AFF]/20 bg-gradient-to-br from-[#EBF4FF] via-white to-[#F0F5FF] dark:from-[#001529] dark:via-[#0D1117] dark:to-[#050A1A] px-8 py-10 mb-8">
  <div className="pointer-events-none absolute -right-16 -top-16 h-64 w-64 rounded-full bg-[#007AFF]/10 blur-3xl" />

  <div className="relative max-w-2xl">
    <span className="mb-4 inline-block rounded-full border border-[#007AFF]/30 bg-[#007AFF]/10 px-3 py-1 text-xs font-semibold uppercase tracking-widest text-[#007AFF]">
      Braintrust vs. The Alternatives
    </span>

    <h1 className="mb-3 text-4xl font-extrabold tracking-tight text-gray-900 dark:text-white leading-tight">
      Why Braintrust in the<br />Age of AI Agents
    </h1>

    <p className="text-lg leading-relaxed text-gray-600 dark:text-gray-300">
      AI agents are everywhere — Claude Code, OpenClaw, OpenHands, 230k+ GitHub stars. So why would you choose Braintrust?
    </p>

    <p className="mt-3 text-lg font-semibold text-gray-900 dark:text-white">
      The short answer: <span className="text-[#007AFF]">Braintrust is built for teams, not for developers tinkering in a terminal.</span>
    </p>
  </div>
</div>

## The Landscape Today

|                            | Braintrust                   | Claude Code            | OpenClaw                    | OpenHands            |
| -------------------------- | ---------------------------- | ---------------------- | --------------------------- | -------------------- |
| **Who it's for**           | Any team member              | Developers             | Technical power users       | Developers           |
| **Where it runs**          | Google Chat (managed)        | Terminal / IDE         | Your machine (daemon)       | Docker / K8s sandbox |
| **Setup time**             | Under a minute               | 30 seconds             | \~2 hours                   | 15–30 minutes        |
| **Technical skill needed** | None                         | High                   | Medium-High                 | High                 |
| **Agent model**            | Narrow specialists           | General-purpose        | General-purpose             | General-purpose      |
| **Memory**                 | Persistent per expert        | Session-based (resets) | Persistent (local)          | Session-based        |
| **Multi-agent**            | Native (multi-expert spaces) | Agent teams (dev-only) | Sub-agents + agent networks | Single agent         |

***

## Why Teams Choose Braintrust

<CardGroup cols={2}>
  <Card title="It Lives Where Your Team Works" icon="message" color="#007AFF">
    Claude Code lives in a terminal. OpenClaw pipes into messaging apps. Braintrust experts **live inside Google Chat** — where your team already shares updates and makes decisions. No new app, no terminal, no Docker. An agent buried in a terminal gets used by 3 developers. An agent in your team's chat gets used by everyone.
  </Card>

  <Card title="Zero Friction for Non-Technical Users" icon="circle-check" color="#34C759">
    **Claude Code:** Install, authenticate, learn terminal patterns.

    **OpenClaw:** 2+ hours setup, Node 22+, plaintext API keys.

    **OpenHands:** Docker containers, sandbox configuration.

    **Braintrust:** Click **Hire**. Start talking.

    If your goal is AI for your marketing or operations team — setup complexity is a dealbreaker.
  </Card>

  <Card title="Narrow Experts Beat Generic Agents" icon="wand-magic-sparkles" color="#5856D6">
    Claude Code, OpenClaw, and OpenHands are all "one agent that does everything." Braintrust takes the opposite approach: a marketplace of specialists, each purpose-built for a domain. Domain-specific agents require fewer resources, eliminate noise, and consistently deliver higher accuracy on domain tasks than generalists.
  </Card>

  <Card title="Managed Platform vs. Your Laptop" icon="cloud" color="#FF9500">
    OpenClaw and Claude Code run on your machine — when your laptop sleeps, your agent sleeps. OpenClaw also stores API keys in plaintext. Braintrust runs as a managed service. Always available, always isolated.
  </Card>

  <Card title="Multi-Expert Collaboration" icon="users" color="#007AFF">
    Most frameworks give you one agent trying to handle everything. Braintrust lets you add multiple experts to the same Google Chat Space. A research expert, a writing expert, and a data expert — each with their specialty — all visible to the entire team. That's how real teams work.
  </Card>

  <Card title="Extensible Without the Complexity" icon="plug" color="#5856D6">
    Braintrust still connects to everything — Google Workspace, Notion, Supabase, any custom API via MCP. The difference: no config files, no YAML, no code. You configure integrations through a UI, and attach reusable [Agent Skills](https://agentskills.io/home) that add domain expertise and repeatable workflows.
  </Card>
</CardGroup>

***

## When Alternatives Might Be the Better Fit

Braintrust isn't the right choice for every use case. Here's when the alternatives genuinely win:

<AccordionGroup>
  <Accordion title="You need a coding agent" icon="code">
    If your primary need is writing code, debugging, refactoring, and managing git workflows — **Claude Code** and **OpenHands** are purpose-built for this. They understand codebases, run tests, and generate pull requests. Braintrust experts are not coding agents.
  </Accordion>

  <Accordion title="You want full local control and privacy" icon="lock">
    If data sovereignty is non-negotiable and you need everything on your own infrastructure with zero external calls — **OpenClaw** (self-hosted) or **OpenHands** (on-premises) give you that. Braintrust is a managed service.
  </Accordion>

  <Accordion title="You're a developer who prefers terminal workflows" icon="terminal">
    If you live in the terminal and want an AI that meets you there — **Claude Code** is excellent. It integrates directly into VS Code, JetBrains, and terminal sessions. Braintrust is designed for chat-based interaction.
  </Accordion>
</AccordionGroup>

***

## The Bottom Line

<Callout color="#007AFF">
  **Claude Code, OpenClaw, and OpenHands are developer tools.** They're powerful, flexible, and increasingly capable — but they assume technical users, local execution, and individual use.

  **Braintrust is an AI team platform.** It puts specialized experts inside the collaboration tool your team already uses, with zero setup, managed infrastructure, and multi-expert collaboration built in.

  If you're building software, use a coding agent. **If you're running a team, hire an expert.**
</Callout>

<CardGroup cols={2}>
  <Card title="Get Started" icon="rocket" color="#007AFF" href="/get-started/quickstart" arrow={true}>
    Hire your first expert in under a minute.
  </Card>

  <Card title="Browse Experts" icon="magnifying-glass" color="#5856D6" href="https://braintrust.ti.trilogy.com/experts" arrow={true}>
    Explore the expert marketplace.
  </Card>
</CardGroup>