Open the app
telecode

A browser-based remote for your coding agents

An open-source, self-hostable command center for AI coding agents. They run on your hardware; your code never leaves your machine.

AGPL-3.0 · 0 bytes leave your machine · End-to-end encrypted · Self-hostable

See it run

Your machine does the work. You drive it from anywhere.

The agent runs on the laptop at your desk. Send a prompt from your phone, watch it work, and approve each real action — over an end-to-end-encrypted relay that only ever sees ciphertext.

Why telecode

Four guarantees, not a feature list

telecode is defined by what it is: open, yours to host, encrypted end to end, and built on real agent sessions. Everything else follows.

  • Open source

    AGPL-3.0

    Built in the open. Read every line, fork it, and own your workflow. No black boxes between you and your agents.

  • Self-hostable

    one command

    Run the whole stack yourself. The relay is a thin multiplexer; your laptop does the work and nothing ever reaches into it.

  • End-to-end encrypted

    ciphertext only

    Prompts, output, and diffs are encrypted in your browser. The relay only forwards ciphertext plus routing metadata. It cannot read your work.

  • Real agent sessions

    tool approvals

    Real agent sessions with real tool approvals, not a screen-scraped terminal. Every consequential action waits for your decision.

How it works

From zero to a steered agent in three steps

No inbound ports, no tunnels to configure. The daemon dials out; you stay in control from any browser.

  1. Install on your machine

    Run one command where your agents should work. The daemon dials out to the relay, so nothing ever reaches into your laptop.

    npx @telecode/cli

  2. Pair from the browser

    The daemon prints a short code. Sign in to the dashboard, enter the code, and this machine is bound to your account.

    telecode pair 7F2-K9Q

  3. Launch and steer

    Pick a device, point at a repo, and send a prompt. Watch output stream and approve each consequential tool call, from your phone or laptop.

Open the app Read the docs

Security & trust

The relay never sees your code

telecode is built so the only thing that has to be trusted is software you can read and host yourself. Everything sensitive stays on your machine.

  • Encrypted before it leaves the page

    Prompts, agent output, and diffs are sealed in your browser. The relay forwards ciphertext plus the routing metadata it needs, and nothing else.

  • Your keys never leave your machine

    Your API keys and your source live where the daemon runs. telecode connects you to them; it never holds copies.

  • Every consequential action is gated

    Each tool call that writes, deletes, or runs waits for an explicit approval, from you, on whatever device you have.

  • Self-host the relay, or read it

    Run the relay on your own infrastructure for full control, or audit the AGPL-3.0 source. There is no proprietary core to trust.

  • Read the security & privacy model

Use cases

Your dev machine, wherever you are

The agent runs where your code, tools, and keys already live. You just decide what it does next, from a laptop, a phone, or a tablet.

  • Review and merge from your phone

    Kick off a code review, read the diff, and approve the merge while you are away from your desk.

  • Start a long task on the commute

    Send a prompt, let the agent grind on your workstation, and check back when the work is done.

  • Triage incidents from anywhere

    Point an agent at a failing service, watch it work, and approve each fix as it goes.

  • Drive every machine at once

    One dashboard across every paired laptop and server. Pick exactly where each job runs.

  • Pair, don’t wait

    Reasoning and output stream live. Steer the agent mid-task instead of waiting for a final answer.

  • Keep regulated code in-house

    Nothing third-party sits in the path. Self-host the relay and the code never touches infrastructure you don’t own.

FAQ

Questions, answered

  • A command center for AI coding agents. The agents run on your own machine; telecode lets you launch, watch, and approve their work from any browser, phone or laptop, without your code ever leaving your hardware.

  • No. Your source and your API keys live where the daemon runs. Everything that does travel (prompts, output, diffs) is end-to-end encrypted in your browser, and the relay only ever sees ciphertext plus routing metadata.

  • Today telecode runs Claude Code agents through the Claude Agent SDK, with real sessions and real tool approvals. Support for more coding agents, such as OpenAI Codex, is on the roadmap.

  • Yes. The agent does its work on your machine, so the daemon needs to be running. That is also the point: the compute, the code, and the keys stay yours.

  • Run the CLI on the machine you want to control. It prints a short pairing code; enter that code in the dashboard and the machine is bound to your account. No inbound ports, no tunnels; the daemon dials out.

  • Yes. telecode is open source under AGPL-3.0. You can self-host the entire stack, including the relay, on infrastructure you own, or read every line first.

Still curious? The documentation and source have the rest.

Your machine. Your keys. Your agents.

Put your agents to work, on your own turf

Open the dashboard and pair your first machine in under a minute, or read every line first. It’s all open source.

Open the app View on GitHub