
Self-hosted remote approvals for Claude Code and Codex: allow tool calls and follow runs from Telegram, Feishu or a web terminal.
by y49· CommunicationUnclaimed · Yours? Claim it
4 images
About
tlive is a self-hosted layer that lets you watch and steer your Claude Code and Codex sessions from your phone. Your agent keeps running in your terminal as usual; tlive uses the agents' own hook and app-server mechanisms to push status and approval cards to Telegram or Feishu, and serves a web dashboard and a real terminal from your own machine. It works with any subscription or API key, and your sessions and data stay on your machine. What you can do with it - Get notified in Telegram, Feishu or a desktop toast when a run finishes, fails or is waiting for you - Turn on remote approval (tlive mode full) to Allow, Deny or Always-allow a tool call from your phone, in parallel with the terminal dialog: whichever answers first wins - Reply to an IM message to continue a stopped run, and answer the agent's questions from IM or the dashboard - Wrap a session (tlive run claude) for a live web terminal you can type into from any device, and send it photos or files - See every session on one dashboard, with its status and last message How it connects - Install the engine with npm install -g tlive, then run tlive setup. It registers the tlive plugin through Claude Code's and Codex's own plugin managers and asks for your IM bot credentials (or skip them and ask your agent to help configure tlive). - The Claude Code plugin from this page adds tlive's hooks, a tlive skill and the /tlive:url and /tlive:status commands. It still needs the tlive engine installed to work. - Codex sessions are watched through an app-server companion; Codex needs no hooks. Good to know - The default posture is notify: tlive watches and notifies but never holds a tool call. Remote approval is opt-in, and tlive mode off turns every hook off. - Nothing is ever auto-denied: if no one answers remotely, your local prompt stays in charge, and your own permissions.deny settings always win. - The dashboard binds to your LAN by default, guarded by a single token; set web.bind to 127.0.0.1 for loopback only. Cards never carry the dashboard link. - /trust on and Always-allow auto-approve until the daemon restarts, so use them with care. - Version 2 is a rewrite of the earlier bridge, with no migration. A community project by y49, MIT licensed.
Once it’s added
“help me configure tlive”
“Connect tlive to my Telegram bot”
“Turn on remote approval so I can allow tool calls from my phone”
“Print the link to this session's web terminal”
“Why didn't I get an approval card in Feishu?”
“Show tlive's status and which mode it's in”
Before you add it
Reads only
Your Claude Code and Codex session events, through their hooks and app-server · tlive's config, daemon status and logs on your machine
Makes changes · your agent can ask first
Sends notifications and approval cards to your Telegram or Feishu chat · Answers tool-call approvals you make remotely, and types IM replies into a wrapped session
Each with its own page
From people who use it
Step by step
Nothing new reaches your agent until a person has looked at it.
No reviews yet. Used tlive? Say how it went.
Write a reviewRuns on your machine.
Runs on your machine.
/plugin marketplace add y49/tlive /plugin install tlive@tlive
Scanned and passed · checked 47 min ago
Plugins people install alongside tlive.
Scanned on every change, and again when our rules improve.
Tell us and a person on the trust team will look at it. Or email [email protected].
Nobody has shared how they use tlive yet.
Security scan
Passed
Safe to install in every agent it supports. Claude plugins.
Plugins people install alongside tlive.
Two paired skills for email triage: teach it how you handle mail once, then it sorts new mail and drafts replies that it never sends.
Lets your agent search and read your Gmail or Outlook mail, save drafts, manage Inbox Zero's automation rules and check your email stats.