Skip to content

Your coding agents have a phone number.

Start, steer and approve Claude Code and Codex from iMessage while they keep working on your Mac.

It asks before anything risky.

Reads and tests run on their own. Migrations, pushes and deploys stop and wait for one word from you.

Both agents, every project, one thread.

Claude Code and Codex, as many repos as you register, and a single conversation to run them all.

It runs on your Mac, not ours.

One outbound connection, nine typed commands, keys in your Keychain. The bridge is open source.

What it can do on your Mac, and what it can’t.

The bridge is the only thing Pagr installs, and every line of it is public. The cloud can ask it for nine things. Nothing else exists to be asked.

Typed commands only
The cloud can send exactly nine command types. Each one is a schema the bridge validates before it runs, and the list is in the public repo.
No remote shell
There is no shell.exec, no arbitrary file read or write, no process spawn. If a capability is not in the schema, the bridge cannot be asked to do it.
Code and credentials stay local
Pagr does not need to upload or persist a copy of your repository in its own cloud. Your Claude and Codex logins never leave your Mac; the bridge does not even read them.
Ed25519 device keys in Keychain
Pairing generates a keypair on your Mac. The private key lives in macOS Keychain. Every command the cloud sends is signed, bound to your device, expires, and cannot be replayed.
Outbound only
The bridge opens one outbound TLS WebSocket to Pagr's gateway. It listens on no ports.
Open source, Apache-2.0
Everything that runs on your machine is public. Read it, build it yourself, and pin the version you trust.

Set up in three steps, then leave the desk.

  1. Install the bridge

    One command. It generates a device key in your Keychain and opens nothing inbound.

    npm i -g @pagr/cli
  2. Pair your Mac

    Confirm it in the browser once. Claude Code and Codex are detected on their own.

    pagr connect
  3. Text it

    Say what you want in plain words. No slash commands, no session ids.

    start claude on checkout-api and fix the failing tests

One plan. Try it free for a week.

7 days free, then $29/mo. Cancel anytime. A card is collected at signup and nothing is charged until the trial ends. Cancel in the dashboard before then and you are not billed.

Pagr

$29per month

Everything included, one price, billed monthly.

  • Claude Code and Codex
  • Natural-language iMessage control
  • A text the moment an agent needs you
  • Approvals with risk tiers
  • Start, steer, and stop sessions
  • Cross-agent review workflows
  • Screenshot attachments
  • The same controls from Claude or ChatGPT
Start free

Cancel anytime in the dashboard. Prices in USD.

Questions

How does the free trial work?

7 days free, then $29/mo. Cancel anytime. You add your payment details when you sign up, nothing is charged during the 7 days, and the first charge happens when the trial ends. Cancel any time before then in the dashboard and you are not billed.

Do I need my computer online?

Yes — Claude Code and Codex run on your Mac, so it needs to be awake and connected. If it is not, supported commands queue for a limited time and run when the paired Mac reconnects.

Does Pagr upload my repository?

The hosted service is designed so it never needs a copy of your repo. The open-source bridge talks to the coding runtimes locally and reports back typed events — status, questions, summaries — not files. Claude and Codex themselves process code according to their own terms.

Do you have my Claude or ChatGPT credentials?

No. Codex authentication stays with the local Codex runtime. Claude support uses a provider-approved integration path; any local API-key fallback stays on your machine. The bridge does not read or transmit provider credentials.

Can it approve commands automatically?

Low-risk automation is configurable. Risky operations always wait for you, and the highest-risk ones — production hints, secrets, pushes to protected branches — arrive flagged as production-level and are answered by the same reply. That is the setting “Approve high-risk actions by text”, which is on by default; turn it off and approving one takes a signed link and a fresh sign-in instead. Denying always works by text.

What happens if I cancel?

The hosted orchestration stops at the end of the paid period. Your local agents, sessions, and repositories are yours and untouched. You can uninstall the bridge with one command.