Live remote MCP
YT2Text MCP Server
Connect Claude, OpenClaw, Hermes, Codex, Claude Code, or your own agent runtime to YouTube transcript and summary tools with one YT2Text API key.
Endpoint
mcp.yt2text.cc/mcp
Auth
Bearer sk_...
Status
Live
Setup
Your key, your quota, your tools.
MCP is not a shared magic account. The remote server receives the user's API key per request, forwards it to the YT2Text API, and applies the same plan limits as REST.
1
Create your API key
Sign in, choose Plus or Pro, then generate a key in the dashboard. The MCP server uses your own YT2Text API key.
2
Add the MCP server
Use https://mcp.yt2text.cc/mcp as the remote MCP URL and send Authorization: Bearer sk_... with each request.
3
Call video tools
Your client chooses when to call tools. YT2Text exposes schemas and descriptions; the LLM decides based on the user request.
Clients
Works with MCP-capable agent environments
Claude
Remote HTTP MCP with a dashboard API key in the Authorization header.
Use the remote server URL and Bearer header.
OpenClaw
Connect as a remote Streamable HTTP MCP server and keep the key in your agent secrets.
Set endpoint + Authorization header.
Hermes
Use the same MCP endpoint from orchestration flows that can call remote MCP tools.
Store YT2TEXT_API_KEY server-side.
Codex
Use YT2Text as an external MCP/tool surface for YouTube transcript and summary work.
Keep credentials outside prompts and repo files.
Claude Code
Use the local stdio adapter when your client prefers command-based MCP servers.
Run the packaged Node adapter.
Custom agents
Any MCP-capable runtime can call the five YT2Text tools over HTTPS.
Forward a user-owned API key per request.
Tool behavior
The LLM decides when to call. YT2Text defines what is callable.
You do not hard-code reactions for every model. The MCP server publishes a small, stable tool surface. The client LLM sees tool names, descriptions, and schemas, then calls the right tool when the user's task needs YouTube text.
| Tool | Purpose | Best use |
|---|---|---|
| process_youtube_video | Queues a YouTube video and returns a job id. | Use for async workflows, long videos, dashboards, and automations. |
| summarize_youtube_video | Queues a video, polls until complete, then returns the result. | Use when the agent should finish the whole transcript + summary flow in one call. |
| get_video_job_status | Checks whether a queued job is pending, processing, completed, or failed. | Use after process_youtube_video before fetching results. |
| get_video_result | Fetches the transcript, summaries, metadata, and exports for a completed job. | Use when the job is completed or when a user asks for the saved output. |
| export_video_markdown | Converts a completed YT2Text result into agent-friendly Markdown. | Use when the next step is notes, docs, research, or publishing. |
Remote MCP
Recommended for hosted clients
{
"mcpServers": {
"yt2text": {
"url": "https://mcp.yt2text.cc/mcp",
"headers": {
"Authorization": "Bearer sk_..."
}
}
}
}Local stdio adapter
Useful for command-based MCP clients
{
"mcpServers": {
"yt2text": {
"command": "node",
"args": [
"/absolute/path/to/yt2text/mcp/yt2text-mcp/dist/index.js"
],
"env": {
"YT2TEXT_API_KEY": "sk_...",
"YT2TEXT_API_BASE": "https://api.yt2text.cc"
}
}
}
}FAQ
Operational answers before you connect a client
Do users configure MCP with their own API key?
Yes. Each user creates a YT2Text API key in the dashboard and sends it as Authorization: Bearer sk_... to the MCP server. The remote MCP server does not store user keys.
Do we need to define how it reacts to every LLM?
No. MCP exposes tools, names, descriptions, and JSON schemas. Claude, OpenClaw, Hermes, Codex, or another client decides when to call a tool. We influence behavior with clear tool descriptions, examples, and safe auth rules.
Does the MCP server call an LLM directly?
No. The MCP server is a tool layer. It forwards authenticated requests to the YT2Text public API, which handles transcripts, summaries, exports, quotas, and plan gates.
Which plan is required?
Plus is the minimum plan for API keys and MCP usage. Pro unlocks batch processing, webhooks, custom prompts, PDF export, infographics, AI chat, and priority queue.