LLMCOM

Your phone. Your coding conversations. One channel.

Connect Claude or ChatGPT to exchange messages with Claude Code or Codex already running on your Mac.

Mac setup · Claude and OpenAI custom connectors
1 · On your Mac

Install and open your setup guide

Paste this into Terminal. It installs LLMCom and its prerequisites, creates your private workspace if needed, and opens a personal setup page.

curl -fsSL https://llmcom-connector.fly.dev/install.sh | sh

Already installed? Run llmcom phone myphone. To update an older version, run the install command above. Existing workspace connections are preserved. The installer supports macOS; allow several minutes for the first download.

Inspect the installer · Source code · PyPI package

ChatGPT / Codex

Add LLMCom to your OpenAI account

On your Mac, install and open the OpenAI setup:

curl -fsSL https://llmcom-connector.fly.dev/install.sh | sh -s -- --client openai

Already updated? Run llmcom phone myphone --client openai.

  1. Open ChatGPT → Plugins, choose Add → Add custom MCP server, and name it LLMCom.
  2. Server URL: https://llmcom-connector.fly.dev/mcp. Select OAuth. Discovery selects Dynamic Client Registration and the llmcom scope automatically; no client secret is needed.
  3. Create the plugin and continue to LLMCom. Copy the one-time pairing code from the private page opened on your Mac and paste it into the LLMCom sign-in page. It expires after 10 minutes.
  4. Install/select LLMCom in your conversation, then ask it to join myphone, say hello, and listen for replies.

Each person pairs their own installation. The shared URL contains no private workspace key. David runs this on his own Mac and links his own OpenAI account.

This is a custom plugin, not a public directory listing. Verify that LLMCom appears in your exact phone Codex/voice chat; availability can depend on the client and workspace. A working desktop connection alone does not prove phone voice support.

2 · In Claude on your desktop

Add your personal connector

The page opened by the command contains your own server URL and credential. Keep it private; don't use a friend's values.

  1. Open Claude → Customize → Connectors. Choose Add custom connector.
  2. Name it LLMCom Remote. Copy the server URL from your personal setup page.
  3. Choose No sign-in. A sign-in warning is expected because this connector uses an API key.
  4. Under Request headers: Header name = Authorization. Value = the complete Bearer … value copied from your personal page. Leave Required checked.
  5. Save. Use the same Claude account on your phone.

This is a custom connector you add by URL. It is not an Anthropic-verified directory listing.

3 · In your existing coding chat

Join the Mac side

Ask your running Claude Code or Codex conversation:

Run llmcom join myphone in this conversation.

Run it inside the coding conversation so replies reach that conversation. Joining from an unrelated Terminal window doesn't identify the chat you want.

4 · On your phone

Say hello, then listen

Start a new Claude chat, enable LLMCom Remote in its Connectors menu, and say or type:

Use LLMCom Remote to join myphone as my-phone, say hello, then listen for replies.

Approve the tools when Claude asks. Ask your coding conversation to reply with a distinctive word, then check the phone receives it.

Listening lasts up to 18 seconds per call. An arriving message resumes that active call. It does not wake an idle conversation or keep listening forever. Ask it to listen again for another window. Keep the Mac awake and online.

When something gets stuck

Claude cannot find the listening tool

Start a new phone conversation after connecting. Where available, use the connector’s refresh-tools control on desktop. An older conversation may retain the previous tool list. Confirm that Llmcom wait appears.

Claude says nobody replied

Ask “Fetch fresh messages from myphone now.” An answer from memory is not a fresh check. When expecting an incoming message, ask “Listen on myphone for 18 seconds.”

The computer is offline or setup stopped

Wake the Mac and run llmcom connector status. Run llmcom phone myphone again to repair/restart the connection and reopen instructions. Saved credentials are reused; no new connector is needed. If the command isn't on your PATH, use ~/bin/llmcom phone myphone.

I want a different room

Run llmcom phone another-room, then join that same room from your coding conversation and phone. The phone command preserves previously enabled rooms.

Does David need access to my computer?

No. Each installation has its own authenticated route and credentials. David's fresh installation creates his own workspace. A room named myphone on his workspace is separate from yours. Sharing a workspace is a separate invitation setup; this page connects a person's phone to their own coding chats.

What runs in the cloud?

A shared Fly gateway forwards authenticated tool calls over an outbound connection from your Mac. It opens no inbound port on your laptop, exposes no general shell, and hosts no language model. The gateway handles message contents in transit. Model usage comes from your Claude or OpenAI plan. The Mac helper must remain running.