# Use vid2context in Claude Desktop

Claude Desktop can read any video through the vid2context MCP server: the transcript with timestamps, the metadata and the chapters. No key is needed. Captions come straight from the platform, and when a video has none, speech-to-text falls back through free and keyless Whisper rungs.

## Install

These steps use a clone of the repo (the npm package is not published yet, so these steps run the server from source).

1. Open Settings, Developer, Edit Config. The file is `~/Library/Application Support/Claude/claude_desktop_config.json` on macOS and `%APPDATA%\Claude\claude_desktop_config.json` on Windows.

2. Clone the repo and install it once (Node.js 22 or newer). Run the next commands from the repo root.

   ```sh
   git clone https://github.com/nirholas/vid2context.git
   cd vid2context
   npm install
   ```

3. Print the config with this clone's paths, then add it to `claude_desktop_config.json`. Merge it into `mcpServers` if the file already has other servers.

   ```sh
   printf '{"mcpServers":{"vid2context":{"command":"%s/node_modules/.bin/tsx","args":["%s/packages/mcp/src/bin.ts"]}}}\n' "$PWD" "$PWD"
   ```

4. Keys are optional. To unlock more free rungs, add an `env` object to the entry, for example:

   ```json
   {
     "mcpServers": {
       "vid2context": {
         "command": "npx",
         "args": [
           "-y",
           "vid2context-mcp"
         ],
         "env": {
           "GROQ_API_KEY": "your-key"
         }
       }
     }
   }
   ```

5. Quit and reopen Claude Desktop so it starts the server.

## Check that it works

The tools menu in a new chat lists vid2context with its tools.

## Try it

Ask Claude Desktop:

> Get the transcript of https://www.youtube.com/watch?v=jNQXAC9IVRw and summarize it with timestamps.

The `get_transcript` tool hands the agent a short header and the timed transcript. This is the real text it returns for "Me at the zoo", rendered by vid2context from a recorded run:

```text
Title: Me at the zoo
Channel: jawed
URL: https://www.youtube.com/watch?v=jNQXAC9IVRw
Duration: 0:19
Transcript: creator-captions, en (youtube-captions)
Chapters: 0:00 Intro | 0:05 The cool thing | 0:17 End

[0:01] All right, so here we are, in front of the elephants
[0:05] the cool thing about these guys is that they have really...
[0:07] really really long trunks
[0:12] and that's cool
[0:14] (baaaaaaaaaaahhh!!)
[0:16] and that's pretty much all there is to say
```

The agent can now cite moments by timestamp. For long videos it can call `get_outline` first and read only the sections it needs.

## Optional keys

Every key is optional. Each one unlocks more free rungs, which run before the keyless ones:

- `GROQ_API_KEY`: Groq free tier: fast Whisper speech-to-text and LLMs
- `GEMINI_API_KEY`: Google AI Studio free tier LLMs
- `OPENAI_API_KEY`: OpenAI, used last and only when set
- `OLLAMA_BASE_URL`: Local Ollama server, for example http://localhost:11434/v1
- `LOCAL_STT_BASE_URL`: Local OpenAI-compatible speech-to-text server, for example http://localhost:8000/v1

The full list, with every model override, is in [Providers and rungs](https://vid2context.com/docs/providers.md).

## Links

- [Claude Desktop MCP documentation](https://modelcontextprotocol.io/docs/develop/connect-local-servers)
- [YouTube transcript MCP server](https://vid2context.com/docs/how-to/youtube-transcript-mcp-server.md): what the tools do
- [Source on GitHub](https://github.com/nirholas/vid2context)