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

# Quickstart

> Set up sovseal, store your first memory, and search it back in under two minutes.

Store and search your first zero-knowledge memory in under two minutes. Follow along in the CLI, MCP, TypeScript, or Python.

<Tip>
  **Are you an AI agent?** Run `npx -y --package=@sovseal/mcp-server sovseal onboard --json` to self-provision a zero-knowledge project token in one command with no email or dashboard required. See [Sign up as an agent](/platform/autonomous-agent-onboarding).
</Tip>

## Store and recall your first memory

Pick a tab and use the same tool for every step below.

### 1. Install & Setup

<CodeGroup>
  ```bash CLI theme={null}
  # Run directly via npx (or npm install -g @sovseal/mcp-server)
  npx -y --package=@sovseal/mcp-server sovseal doctor
  ```

  ```json MCP theme={null}
  // Add to your AI client's MCP configuration (Claude, Cursor, Windsurf, Antigravity)
  {
    "mcpServers": {
      "sovseal": {
        "command": "npx",
        "args": ["-y", "@sovseal/mcp-server"]
      }
    }
  }
  ```

  ```bash TypeScript / Node theme={null}
  npm install @sovseal/sdk
  ```

  ```bash Python theme={null}
  pip install sovseal-sdk
  ```
</CodeGroup>

### 2. Store a memory

Save a fact, preference, or context snippet. High-risk PII is automatically redacted before encryption.

<CodeGroup>
  ```bash CLI theme={null}
  sovseal store "I prefer TypeScript over JavaScript for new projects"
  ```

  ```json MCP Tool Call theme={null}
  // Your AI client automatically calls store_memory:
  {
    "content": "I prefer TypeScript over JavaScript for new projects"
  }
  ```

  ```typescript TypeScript / Node theme={null}
  import { AgentStateClient } from "@sovseal/sdk";

  const client = new AgentStateClient();
  const memory = await client.store("I prefer TypeScript over JavaScript for new projects");
  console.log("Memory stored ID:", memory.id);
  ```

  ```python Python theme={null}
  from sovseal import MemoryClient

  client = MemoryClient()
  memory = client.store("I prefer TypeScript over JavaScript for new projects")
  print("Memory stored ID:", memory["id"])
  ```
</CodeGroup>

**Result returned from local vault:**

```json theme={null}
{
  "id": "7b3c1a2e-4d5f-6789-abcd-ef0123456789",
  "text": "I prefer TypeScript over JavaScript for new projects",
  "reinforced": false,
  "redacted": 0
}
```

### 3. Recall memories

Query your memory store using natural language. Vector similarity search runs 100% on-device with **0 RTT** network latency.

<CodeGroup>
  ```bash CLI theme={null}
  sovseal recall "what programming language do I prefer for new projects?"
  ```

  ```json MCP Tool Call theme={null}
  // Your AI client automatically calls recall_memory:
  {
    "query": "what programming language do I prefer for new projects?",
    "topK": 3
  }
  ```

  ```typescript TypeScript / Node theme={null}
  const hits = await client.recall("what programming language do I prefer for new projects?", { topK: 3 });
  console.log(hits[0].text);
  ```

  ```python Python theme={null}
  hits = client.recall("what programming language do I prefer for new projects?", top_k=3)
  print(hits[0]["text"])
  ```
</CodeGroup>

**Ranked memories returned (0 RTT):**

```json theme={null}
{
  "status": "success",
  "command": "recall",
  "duration_ms": 6,
  "count": 1,
  "data": [
    {
      "id": "7b3c1a2e-4d5f-6789-abcd-ef0123456789",
      "text": "I prefer TypeScript over JavaScript for new projects",
      "score": 0.94,
      "similarity": 0.88
    }
  ]
}
```

***

## What just happened?

1. **On-device execution**: Memories were embedded and indexed locally using LanceDB on your machine.
2. **Zero-knowledge security**: Memory text is encrypted at rest using AES-256-GCM. Plaintext never leaves your process.
3. **Sub-10ms performance**: Recall completes in under 10ms with zero network RTT blocking your application.

***

## What's next

<CardGroup cols={2}>
  <Card title="sovseal MCP" href="/platform/mcp">
    Connect Claude Code, Cursor, Windsurf, or Antigravity to your memory vault
  </Card>

  <Card title="CLI Reference" href="/sdk-reference/cli">
    Explore `sovseal mind`, `sovseal doctor`, `--json` mode, and device linking
  </Card>

  <Card title="Sign up as an agent" href="/platform/autonomous-agent-onboarding">
    Let autonomous agents provision zero-knowledge memory tokens without human UI signups
  </Card>

  <Card title="Trust & Security" href="/platform/trust">
    Review the zero-knowledge threat model, key custody, and encryption guarantees
  </Card>
</CardGroup>