vibelive MCP Server
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@vibelive MCP Serverstart a shared claude code session and show the join URL"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
vibelive
Multiplayer for agentic coding terminals — shared Claude Code/Codex/Gemini sessions with live Figma-style cursors, presence, and in-terminal chat.
Part of the Vibe Suite — companion tools for agentic coding CLIs (Claude Code, Codex, Gemini, Grok/pi, Kimi). Ships as CLI + npm package + MCP server.
Local-first: runs on your own machine. v0 is a host-authoritative local/LAN session: one machine hosts a wrapped agent and owns the truth; peers join over a local WebSocket. Full end-to-end-encrypted relay fan-out to ~1,000 participants is on the roadmap (see docs/tech-spec.md). Session sharing is gated by the consent ledger in @pooriaarab/vibe-core (scope share:session), and is revocable.
Install
npm install -g vibelive
# or use it as a library / MCP server without the global install:
npm install vibeliveRequires Node ≥ 18.
Related MCP server: Claude Relay
Quick start
Share a wrapped agent (anything you'd run in a terminal) and print a join URL:
vibelive host -- claude # share a Claude Code session
vibelive host -- python -i # share a REPL
vibelive host --port 4474 --name ada -- claudeThe host prints something like:
vibelive host ready — sharing claude
join: ws://localhost:54157
lan: ws://10.0.0.179:54157
you are the driver. /release to hand off, /drive to take back, /quit to end.From another terminal (or another machine on the LAN):
vibelive join ws://localhost:54157 --name adaThe joiner sees the live agent output plus a presence/chat line. Slash commands while joined:
/drive— request the write token (queued FIFO behind other drivers)/release— relinquish the token/type <text>— send input to the wrapped agent (driver only)/quit— leave the session
Everything else you type is sent as chat.
Who can write to the agent?
Exactly one participant — the driver — holds the write token at any time (see src/arbitration.ts). Everyone else always has read + chat + cursor; only agent-write is arbitrated, so two people never interleave garbage into one stdin. The host user starts as the driver.
MCP server
Run vibelive as a Model Context Protocol server over stdio:
vibelive mcpIt exposes two tools an agent can call:
tool | description |
| Start a host+relay wrapping a command; returns the |
| List active sessions (id, url, participants, current driver). |
Example (Claude Code / any MCP client config):
{
"mcpServers": {
"vibelive": { "command": "vibelive", "args": ["mcp"] }
}
}Library
npm install vibeliveimport { createHost, createRelay, joinSession, WriteArbiter } from 'vibelive';
// Host-authoritative session on an ephemeral port.
const host = createHost({ command: ['claude'] });
const relay = await createRelay({ port: 0, hostHandle: host, initialDriver: 'host' });
console.log(relay.url); // ws://localhost:<port>
const client = joinSession({ url: relay.url, name: 'ada' });
client.onOutput((text) => process.stdout.write(text));
client.requestControl(); // ask to driveHow it works (v0)
Three channels over one WebSocket, each with the guarantees its data needs (details in docs/tech-spec.md):
Agent output — ordered, sequence-numbered append-log; the host is the sole author; late joiners get snapshot + tail.
Presence / cursors — ephemeral, high-frequency, lossy is fine (coalesced/interpolated client-side; the relay forwards for v0).
Chat + control — reliable, ordered; agent-write is mediated by the
WriteArbiterso there is never more than one driver.
The correctness-critical piece is src/arbitration.ts — a pure, fully unit-tested state machine enforcing "never two concurrent agent-writers," FIFO turn-taking, and "release when not driver is a no-op."
Prototype
Interactive, self-contained UX prototype (no build, no network): open docs/prototype.html in a browser.
Specs
docs/tech-spec.md— scale (10/100/1000), transport tiers, tests.docs/vibe-core-spec.md— shared suite spine (cascade, hooks, consent).
Roadmap
v0 (this release): host-authoritative local/LAN multiplayer — host + relay + arbitration + CLI + MCP, all over plain pipes/WebSocket.
Next: real PTY wrapping (
node-pty) for full TTY semantics (resize, raw-mode programs, signals), cursor interpolation, and a richer terminal renderer.Scale: the dumb, e2e-encrypted, self-hostable relay (pub/sub fan-out to ~1,000, relay reads only ciphertext) — see tech-spec §2 "Large" tier.
License
MIT
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- Alicense-qualityDmaintenanceEnables AI agents to interact with terminal environments through multiple concurrent PTY sessions. Supports cross-platform terminal operations including command execution, session management, and real-time communication.103MIT
- Alicense-qualityBmaintenanceEnables real-time communication between Claude Code instances across multiple machines via WebSocket, allowing context sharing, task handoffs, and coordination between sessions.883MIT
- Alicense-qualityBmaintenanceEnables AI agents to execute commands in a shared tmux session that is simultaneously visible to a human via a web-based terminal UI.1MIT
- Alicense-qualityDmaintenanceEnables AI agents to spawn and interact with real terminal sessions, capturing screenshots of rendered TUI output and sharing live sessions for debugging.41MIT
Related MCP Connectors
The team layer for AI coding agents: shared contracts, collision alerts, E2EE sessions.
Agent-native collaboration network: orchestrate a team of long-running agents from any MCP client.
Build, validate, and deploy multi-agent AI solutions from any AI environment.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/pooriaarab/vibelive'
If you have feedback or need assistance with the MCP directory API, please join our Discord server