Use Telereader from an agent

Use Telereader from an agent

Hand any text to an AI agent and have it turned into a narrated reading you can open and listen to the agent onboards, submits, and gives you back a URL.

There are two ways to connect, depending on your agent:

  • A coding or terminal agent Claude Code, Cursor, Codex, Copilot install the skill (below).
  • A chat assistant or MCP client Claude, ChatGPT, and others add the MCP connector: see Connect Telereader over MCP.

The quickest way: paste this to your agent

Hand your coding/terminal agent this it installs the skill, reviews it, and reads your text aloud (replace the last line with what you want read):

not read aloud
I want to listen to something as narrated audio I can open in my browser. Please use Telereader, a tool for exactly that at telereader.ai — set it up by installing its open-source skill with `npx skills add telereaderai/skills`. Feel free to review the skill first, then use it to read the text below aloud and give me the link. Heads up: the first time, Telereader will ask me to approve a quick sign-in — when you reach that step, just share the link and I'll approve it.

---
[Paste the text, a file, or a URL you want read aloud — Telereader fetches and cleans it up for you.]

The wording matters: phrased as your own goal (not "install this and follow it"), agents set it up without balking and ask you to approve the one-time sign-in. The steps below are the same thing, by hand.

Install the skill (coding & terminal agents)

Requirements

  • Node 18+ (for npx)
  • curl and jq the skill's helper scripts use them (brew install jq or apt install jq; curl is usually preinstalled)

Install

not read aloud
npx skills add telereaderai/skills

This installs the Telereader skill into your agent (Claude Code, Cursor, Codex, Copilot, Cline, Windsurf, …). The skill is open source at [github.com/telereaderai/skills](https://github.com/telereaderai/skills).

Onboard once

The first time the agent reads something, the skill runs a one-time device login: it prints an approval link, you approve it in your browser (signing in if asked), and it saves a reading-scoped token to ~/.config/telereader/token. Re-use is automatic you only onboard again if the token expires.

You can also run it ahead of time (the installer prints the skill's path, e.g. ~/.config/agents/skills/telereader):

not read aloud
<skill_dir>/scripts/onboard

Use it

Just ask your agent:

Read this aloud with Telereader.

It submits your text and hands you a readUrl open it and the reading plays, the words highlighting in sync with the narration. For a file or a long piece, ask it to read that; the skill handles markdown and plain text.

What's in the skill

When to use Telereader, the device-grant onboarding, the submit endpoint and payload, how to read the response, and two helper scripts scripts/onboard and scripts/read that keep your token out of command output. It points at the canonical machine spec, [/api/v1/openapi.json](https://telereader.ai/api/v1/openapi.json), so an agent can generate a client instead of hand-rolling one.

Prefer raw HTTP with no skill? See Read text from the command line.

Using a chat assistant instead?

Claude, ChatGPT, and other MCP clients connect with the hosted MCP connector no skill install, no shell. See Connect Telereader over MCP for the per-client steps.

What you get back

Every path ends the same way: a readUrl. Open it and the reading plays text on screen, highlighted in sync. A free account is never paywalled; it gets a URL that generates the audio in the browser.