Your podcasts, inside Claude.
PodPast.ai ships a Model Context Protocol (MCP) server that connects your entire podcast and YouTube vault to Claude Desktop. Add two lines to Claude's config file, reconnect, and Claude can search your transcripts, synthesize answers across episodes, and cite the exact timestamp — all from within any conversation.
How to install in 5 steps
Create a PodPast.ai account and add your first podcast feed
Sign up at podpast.ai and paste any YouTube URL or podcast RSS feed. The first batch of episodes will begin transcribing immediately. Add at least one feed before testing the MCP connection.
Copy your personal MCP URL
After signing up, your unique MCP URL is shown in your account dashboard — it looks like: https://api.podpast.ai/mcp/YOUR_TOKEN. Copy the full URL.
Open Claude Desktop → Settings → Developer → Edit Config
Open Claude Desktop on macOS or Windows. Go to Settings (⌘, on Mac), click the Developer tab, then 'Edit Config'. This opens claude_desktop_config.json.
Add the PodPast MCP server entry
Add this JSON block inside the
mcpServersobject in your config file:{ "mcpServers": { "podpast": { "type": "http", "url": "https://api.podpast.ai/mcp/YOUR_MCP_TOKEN" } } }Replace
YOUR_MCP_TOKENwith the token from your account dashboard (the UUID at the end of your MCP URL).Restart Claude Desktop — your Pod is searchable from any conversation
Quit and reopen Claude Desktop. The hammer icon in the chat toolbar confirms active MCP tools. Test with: 'Search my pod for [any topic from a podcast you added]'.
MCP tools exposed by PodPast.ai
search_pod
Semantic search across your vault. Returns ranked transcript chunks with timestamps, episode name, feed name, and a direct link to the source.
ask_pod
Synthesized answer with inline citations and timestamps. Claude calls this tool, PodPast retrieves relevant chunks, and Claude composes a grounded answer — similar to RAG but natively integrated.
add_to_pod
Add a new YouTube URL or RSS feed mid-conversation. Useful when you encounter a new source during research and want to ingest it without leaving Claude.
get_pod_info
Returns a summary of your vault: total feeds, total episodes, processing status, and Deepgram minute balance. Use it to check what's in your library.
Claude MCP questions
- What is the Model Context Protocol (MCP)?
- The Model Context Protocol is an open standard, developed by Anthropic, that lets external tools expose data and capabilities to Claude (and other LLMs) in a structured way. MCP servers communicate over a local stdio or HTTP transport and register a list of tools that Claude can call mid-conversation. PodPast.ai's MCP server exposes four tools: search_pod, ask_pod, add_to_pod, and get_pod_info.
- Does PodPast.ai work with Claude Code?
- Yes. Claude Code reads the same claude_desktop_config.json as Claude Desktop. If you have the PodPast MCP server configured, you can reference your podcast vault from within Claude Code sessions — useful for research-heavy engineering or documentation tasks where you want to pull in context from technical podcast content.
- Can I use PodPast.ai inside Claude Projects?
- Claude Projects can access MCP servers configured at the account level in Claude Desktop. If your PodPast server is running, it is available within any project conversation. This lets you pre-load a Project with podcast context relevant to a specific research area and then ask synthesis questions across both the Project files and your vault.
- How does PodPast.ai compare to ChatGPT's built-in transcript reading?
- ChatGPT can read a transcript if you paste it directly into the context window — which is bounded by the model's context limit and requires manual effort for every session. PodPast.ai maintains a persistent vector index of your entire library and retrieves only the relevant chunks per query, so you never hit context limits and the retrieval is automatic from within any conversation.
Connect your podcasts to Claude today
Free tier. MCP access included. No credit card required.
Start free