Most of the work around audio is not creative. It is clerical: paste the link, pick the publication, add the story to the right playlist, check that the podcast feed updated, look up what it cost. The Everlit MCP server hands that work to the AI assistant you already have open.
What it is
MCP stands for Model Context Protocol, an open standard that lets an AI assistant use outside tools instead of only talking about them. Everlit runs an MCP server at https://api.everlit.audio/v1/mcp. Connect it once to Claude, Claude Code, Cursor, VS Code or any other MCP client, and your assistant can work inside your Everlit account: the same publications, playlists and voices you use in Studio.
Connecting takes about a minute. Add the address as a connector, sign in with your Everlit Studio login, choose which publisher the connection may act as, and approve. There is no key to copy and nothing to install.
What you can ask for
Then you ask for the work in plain language. A few requests people run today:
- “Publish these five URLs to The Daily, pin them to Morning Briefing in this order, and give me the embed for the lead story.”
- “Make a playlist of everything tagged climate, fill it automatically from now on, turn the podcast feed on, and give me the Apple and Spotify submission links.”
- “Narrate everything our partner’s RSS feed publishes into the Wire publication.”
- “Add these 30 surnames to the pronunciation library, and tell me what we spent on speech this month.”
You do not pick the tools. The assistant does, from 52 of them in 10 groups, then tells you what it did and hands back the links.
More than text-to-speech
It would have been easy to ship a server with one tool that turns text into an MP3. This one exposes what you actually do in Studio:
- Articles: publish from a URL or from text, edit, list and take down, with pauses, music and a second voice set inline.
- Publications: read and change the defaults every future article inherits.
- Playlists: create them, pin and reorder stories, or set the rules they fill themselves by.
- Podcast feeds: fill in the show’s details, switch the feed on, and track where the show stands on Apple Podcasts, Spotify, Pocket Casts, Amazon, YouTube and Overcast.
- Pronunciations, music beds and voices, including cloning a voice with the speaker’s consent on record.
- Speech on its own, with word timings and captions, plus your usage and spend for any date range.
Standing orders
Some requests do not end when the conversation does. Ask your assistant to narrate everything a partner’s RSS feed publishes, and Everlit checks that feed on a schedule and narrates each new item into the publication you chose, from then on. Ask for a playlist of everything tagged “election”, and it keeps filling as you publish. You set the rule once, in a sentence.
Built to be trusted with your account
Handing an assistant your publishing account should make you a little careful, so the server is careful too. Every tool is marked as read-only or as one that changes or deletes something, which means your client can ask you to confirm a delete and never bother you about a list. The connection acts only as the publisher you approved when you signed in. For automated jobs with no browser, such as a nightly script, a static API key works in place of the sign-in.
What it costs
Connecting the server costs nothing. What you do through it is billed exactly as it would be in Studio: articles count against your publication’s monthly plan, and speech is included on every paid Studio plan and pay as you go otherwise. Nothing is metered differently because an agent asked.
How to start
If you already publish with Everlit, you can connect now: the address and the steps for each client are on the Agents and API page. If you do not have an account yet, your assistant can ask for one on your behalf, or you can request access on the same page. A person reads every request, usually within one business day.