# ByteAsk Embedded Docs MCP server

> Lets AI coding agents search embedded and firmware reference docs for verbatim, page-cited snippets on registers, protocols, and standards.

- Repository: https://github.com/ByteAsk/ByteAsk-Embedded-MCP
- Page: https://hostedmcp.org/servers/byteask/byteask-embedded-mcp
- Transport: remote
- Stars: 26
- License: MIT
- Latest version: 1.0.1

## Install

### Claude Code

```sh
claude mcp add --transport http byteask-embedded-mcp https://mcp.byteask.ai/mcp
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "byteask-embedded-mcp": {
      "url": "https://mcp.byteask.ai/mcp"
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "byteask-embedded-mcp": {
      "type": "http",
      "url": "https://mcp.byteask.ai/mcp"
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"byteask-embedded-mcp","type":"http","url":"https://mcp.byteask.ai/mcp"}'
```

### Codex

```sh
codex mcp add byteask-embedded-mcp --url https://mcp.byteask.ai/mcp
```

### Codex config.toml (`~/.codex/config.toml`)

```toml
[mcp_servers.byteask-embedded-mcp]
url = "https://mcp.byteask.ai/mcp"
```

### Gemini CLI

```sh
gemini mcp add -s user --transport http byteask-embedded-mcp https://mcp.byteask.ai/mcp
```

### Gemini settings.json (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "byteask-embedded-mcp": {
      "url": "https://mcp.byteask.ai/mcp",
      "type": "http"
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `BYTEASK_BACKEND` | no | no |  |
| `MCP_TRANSPORT` | no | no |  |

## FAQ

### What does the ByteAsk Embedded Docs MCP server return?

It returns verbatim source snippets with section and page citations, or an honest no-match response instead of an invented answer.

### Which tools does the server expose?

It exposes search_docs for ranked page-cited results, get_context to expand a hit to its full section, and request_document to ask for a missing document.

### Can I run the server locally without API keys?

Yes, the bundled SampleBackend serves illustrative public-knowledge records so search_docs works immediately with no keys.

### How do I plug in my own retrieval engine?

Implement the two-method SearchBackend interface, expose a make_backend factory, and set the BYTEASK_BACKEND environment variable.