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

# Quickstart Overview

> Choose the right integration path for your AI agents

# Pick your path

<Note>
  These quickstarts are available in [Spanish (Español)](/es/quickstart/overview) and [French (Français)](/fr/quickstart/overview) — six pages per language have been translated.
</Note>

## Create your account

Sign up at [mnemom.ai/claim](https://mnemom.ai/claim) with your **email and password** — that's the complete form. No company field, no CRM sync. At signup, a [personal organization](/concepts/personal-organization) is provisioned automatically and linked to your account. See the [Sign up reference](/api-reference/endpoint/post-auth-sign-up) for the API equivalent.

## Install and authenticate

Fastest path — install the CLI and route your first request:

```bash theme={null}
npm install -g @mnemom/mnemom && mnemom login
```

That authenticates the CLI against your Mnemom account and configures local tracing against `gateway.mnemom.ai`. Your agent is registered automatically the first time you route a request through the gateway. No code changes, no API key plumbing. You'll claim the auto-created agent once to bind it to your account — see the [Gateway quickstart](/quickstart/gateway). Keep reading if you need a deeper integration.

<Tip>
  **Been invited to spar in the Dojo? One command, zero install.** If someone handed you a `/try-me` invite token, you don't need to install anything — run:

  ```bash theme={null}
  npx @mnemom/mnemom@latest try-me <token>
  ```

  It runs the whole flow end-to-end — born → claimed → declared (alignment + protection cards) → sparring — explaining each step as it goes. Use `@latest` (not bare `npx @mnemom/mnemom`) so a returning machine doesn't reuse a stale cached build. A Python equivalent, `uvx mnemom try-me <token>`, is coming.
</Tip>

There are three ways to add alignment verification and integrity checking to your AI agents. Choose the one that fits your constraints.

<Note>
  Every Mnemom agent has **two cards** — an [alignment card](/concepts/agent-cards) (who the agent is and what it's allowed to do) and a [protection card](/concepts/protection-card) (how it's defended). Both compose from platform > org > agent scope. Whichever path you pick below, you'll end up authoring these two cards in YAML; the [card management guide](/guides/card-management) and [protection card management guide](/guides/safe-house-config) cover the lifecycle.
</Note>

<CardGroup cols={3}>
  <Card title="Mnemom Gateway" icon="bolt" href="/quickstart/gateway">
    **Recommended.** Wrap any LLM API. Zero code changes. Full AAP + AIP compliance in minutes. Works with Anthropic, OpenAI, and Gemini.
  </Card>

  <Card title="SDK Direct" icon="code" href="/quickstart/sdk-direct">
    Integrate AAP and AIP directly into your application code. Define alignment cards, generate traces, run integrity checks, and verify coherence with full programmatic control.
  </Card>

  <Card title="Self-Hosted" icon="server" href="/quickstart/self-hosted">
    Run the Mnemom Gateway, the API, and the database on your own infrastructure. Complete data residency. Cloudflare Workers + Supabase stack.
  </Card>
</CardGroup>

## Comparison

| Feature               | Gateway                         | SDK Direct                 | Self-Hosted                     |
| --------------------- | ------------------------------- | -------------------------- | ------------------------------- |
| **Setup time**        | 5 minutes                       | 30 minutes                 | 2 hours                         |
| **Code changes**      | None                            | Moderate                   | Moderate                        |
| **AAP traces**        | Automatic                       | Manual                     | Automatic                       |
| **AIP integrity**     | Automatic                       | Manual                     | Automatic                       |
| **Dashboard**         | Included                        | Included                   | Self-hosted                     |
| **Data residency**    | Mnemom cloud                    | Mnemom cloud               | Your infrastructure             |
| **Enforcement modes** | observe / nudge / enforce       | You implement              | observe / nudge / enforce       |
| **Policy governance** | [CLPI](/concepts/clpi) built-in | You configure              | [CLPI](/concepts/clpi) built-in |
| **Multi-provider**    | Anthropic, OpenAI, Gemini       | Any (you extract thinking) | Anthropic, OpenAI, Gemini       |
| **Cost**              | Usage-based                     | Usage-based                | Free (you pay infra costs)      |

## Which should I choose?

<AccordionGroup>
  <Accordion title="I want the fastest path to compliance">
    Use the **Mnemom Gateway**. Install the CLI, run `mnemom login`, point your LLM client at `gateway.mnemom.ai`, and your API calls are automatically traced and verified. No code changes required.
  </Accordion>

  <Accordion title="I need full control over tracing and verification">
    Use **SDK Direct**. You define exactly when traces are generated, how integrity checks run, and what happens on violations. Best for custom agent frameworks or when you need to integrate with existing observability systems.
  </Accordion>

  <Accordion title="My data cannot leave my infrastructure">
    Use **Self-Hosted**. Deploy the Mnemom Gateway stack (Cloudflare Workers + Supabase) on your own accounts. All data stays in your infrastructure. Same automatic tracing as the gateway, but you control everything.
  </Accordion>

  <Accordion title="I'm building an agent-to-agent system">
    Start with **SDK Direct** for value coherence checking between agents. You can combine this with the gateway for tracing, or handle traces yourself. See the [A2A integration guide](/protocols/aap/a2a-integration) for details.
  </Accordion>
</AccordionGroup>