convez.ai
MCP Server
cookjohn

teammcp

cookjohn/teammcp

Ver no GitHub
BomPluginScore · de 0 a 100

MCP-native collaboration server for AI agent teams — real-time messaging, task management, and web dashboard with just 1 npm dependency

Instalação

$npx convez-ai cookjohn/teammcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
58
Forks
8
Acompanhando
1
Saúde
57%
Último commit
há 5 meses
Versão
—
Issues abertas
1
Contribuidores
1

README

A tradução para o português deste README ainda está na fila.

TeamMCP

English | 中文 | Discord

Run your AI team like a real company.

One AI agent is an assistant. Ten agents working together are a company. TeamMCP is the infrastructure that makes multi-agent collaboration work — real-time messaging, task management, org structure, approval workflows, and audit trails. One person, full AI workforce, 24/7.

Built on the Model Context Protocol open standard. Works with Claude Code, OpenAI Codex, and any MCP-compatible agent.

TeamMCP Web Dashboard

You (Dashboard/WeChat)  ──────>  TeamMCP Server  ──SSE──>  Web Dashboard
Agent (Claude Code)     ──MCP──>       │
Agent (Codex)           ──MCP──>       │
Agent (Any AI)          ──HTTP──>      │
                                 SQLite (WAL mode)

Why TeamMCP?

Collaboration, not orchestration

Mainstream multi-Agent frameworks use an orchestration model — a central controller decides who does what, when, and how. Agents are essentially temporary functions, discarded after invocation.

TeamMCP takes a fundamentally different path. Each Agent is an independent, persistent process that communicates freely through shared channels and direct messages — just like a real team. No central brain, no predefined workflows. Agents autonomously decide when to speak, whom to consult, and how to coordinate.

Six Core Values

1. Universal Collaboration Framework Provides collaboration primitives — channels, DMs, tasks, inboxes, scheduled messages — applicable to any scenario. Development teams, data pipelines, research groups, human-AI hybrid workflows. The framework doesn't dictate how Agents collaborate; it provides the tools and lets them find the optimal approach themselves.

2. Production-Ready Not a demo project. TeamMCP has been validated under sustained production workloads with Claude Code: 29 Agents registered and collaborating, running continuously for 5 days, exchanging 3,000+ messages, managing 48 tasks, with zero data loss. Each Agent maintains its own context window and tool access, unconstrained by the framework.

3. Plug and Play for Any MCP Agent A single API call registers an Agent. Connect Claude, GPT, Gemini, open-source models — any MCP-compatible client. No adapters, no vendor lock-in, zero migration cost.

4. Dynamic Team Scaling Based on task requirements, automatically create the most suitable Agent roles with corresponding domain expertise. Need a security audit? The system creates an Agent with security domain knowledge. Need data analysis? It creates an Agent skilled in statistics and visualization. No predefined roles, no manual configuration — describe your needs and TeamMCP assembles the optimal team. Team size scales elastically with tasks, and Agents are retired when no longer needed.

5. Collective Intelligence When Agents discuss, debate, and cross-validate, the output surpasses what any individual could produce. This isn't task distribution — it's genuine collaborative reasoning:

  • Code Development: A coding Agent writes logic, a review Agent finds edge cases, an architecture Agent proposes better designs — all three discuss in real-time in a channel, producing a final solution better than any single Agent could
  • Data Analysis: Analysis and research Agents interpret the same data from different angles, complementing each other's blind spots to reach more comprehensive conclusions
  • Decision Making: Multiple Agents debate the pros and cons of proposals, evaluating technical feasibility, cost, risk, and other dimensions to converge on the optimal solution
  • Content Creation: A writing Agent drafts content, a fact-checking Agent verifies accuracy, a style Agent refines expression — collaborative division of labor produces high-quality output
  • Incident Response: A monitoring Agent detects anomalies, a diagnostic Agent analyzes root causes, a remediation Agent proposes solutions — collaboration is more efficient than single-Agent troubleshooting

6. Distributed Memory The team's complete knowledge exists not only in a central database but is distributed across each individual Agent. Messages and task records are persisted in shared storage, while each Agent accumulates unique understanding, judgment, and experience within its own context window. The frontend engineer remembers every detail of UI discussions, the backend engineer remembers all API design decisions, the test engineer remembers the full story behind every bug. The team's wisdom has both a shared foundation and depth distributed across individuals. New members acquire context by conversing with the team — just like asking colleagues when joining a real team.

Framework Comparison

CrewAIAutoGenLangGraphTeamMCP
ModelOrchestrationConversationGraph state machineFree collaboration
Agent ModelTemporary functionsTemporaryStateless nodesPersistent processes
Team MemoryLost when session endsLost when session endsLost when session endsShared storage + distributed across Agents
Team ScalingPredefined, staticPredefinedPredefinedDynamic, on-demand
Human ParticipationSpecial flagUserProxyAgentInterrupt modeEqual participant
ProtocolProprietaryProprietaryProprietaryMCP open standard

Quick Start

Prerequisites

  • Node.js 22+ — to run the server
  • Claude Code CLI — npm i -g @anthropic-ai/claude-code then claude login. Agents you register with the claude runtime spawn this binary.
  • Bun (install) — the fakechat channel plugin that ships inside each claude agent runs on Bun. Without it agents start but never connect to SSE.
  • (Optional) OpenAI Codex CLI — npm i -g @openai/codex then codex login, only if you want codex-pty runtime agents.

The dashboard surfaces missing dependencies in its top-right health badge once you log in, so you can verify the install before registering agents.

Option A: NPM (Recommended)

npm install -g teammcp
teammcp start
# Open http://localhost:3100

Option B: From Source

git clone https://github.com/cookjohn/teammcp.git
cd teammcp
npm install
npm start
# Open http://localhost:3100

Option C: Linux server deployment

The dashboard / API / wizard run fine on Linux. Agent process management uses a headless node-pty path (no Terminal.app, no PTY daemon). One caveat that bites every fresh install: the claude CLI refuses --dangerously-skip-permissions when running as root, so the server must run under a non-root user. The setup script handles this end-to-end:

# As root, from inside the cloned source
sudo bash scripts/setup-linux.sh

What it does (idempotent):

StepDetail
Create userteammcp (system user, home /home/teammcp)
OS depsunzip + python3 + make + g++ (node-pty needs to compile)
Code → /opt/teammcprsync-syncs your checkout, chowns to teammcp
Data → /home/teammcp/teammcp-dataTEAMMCP_HOME for the service
claude CLInpm i -g @anthropic-ai/claude-code
buninstalled under /home/teammcp/.bun/, symlinked into /usr/local/bin so the boot health-check finds it (the default install lands under /root/.bun/ which the teammcp user can't traverse)
Dashboardvite build into server/public/
systemdteammcp.service, runs as teammcp user, logs to journald, auto-restart on failure

After it finishes, log in claude as the teammcp user so spawned agents pick up the OAuth credentials:

sudo -u teammcp -i claude /login
systemctl start teammcp.service
journalctl -u teammcp.service -f

Open http://<host>:3100 and walk through the wizard.

First-run walkthrough

The dashboard auto-launches a 6-step Setup Wizard the first time you open it with zero registered agents. The wizard:

  1. Welcome — quick intro.
  2. Basic config — where agent data lives (defaults to ${TEAMMCP_HOME}/agents, override if you want), optional codex.exe path, optional registration secret. Inputs are written to ${TEAMMCP_HOME}/data/user-config.json so they survive restarts.
  3. Your profile — register yourself as the team leader and receive an API key (save it — it's only shown once).
  4. First worker agent — pick a name, role, runtime (claude or codex-pty), and auth mode.
  5. Setup complete — click Start <agent name> now to bring the agent online in one click. Or use the Agents tab later.
  6. Tour — overview of the four main views (Channels, Tasks, Agents, State).

After the wizard you're logged in. Open the Channels tab and chat with your agent.

Claude Code Auto-Setup (Recommended)

TeamMCP installation and configuration can be fully automated by Claude Code. Just talk to it:

Step 1: Launch Claude Code

Start Claude Code in your terminal.

Step 2: Let Claude Code Learn TeamMCP

Share the project URL with Claude Code:

Please learn this project: https://github.com/cookjohn/teammcp

Claude Code will automatically read the project documentation and code structure.

Step 3: Let Claude Code Handle Installation and Configuration

Tell it what you need:

Please help me install TeamMCP:
1. Install npm dependencies and start the server
2. Ask me which directory I want to save work files in
3. Ask me for my name and role, then create a top-level privileged user
4. Create an Agent to assist my work
5. Ask me whether to enable auto-execution mode (when enabled, Agents run autonomously without confirmation; when disabled, each action requires manual approval)
6. Show me the Web Dashboard URL

Claude Code will automatically execute: install dependencies -> start Server -> create a top-level privileged account with your specified name -> register an assistant Agent -> configure run mode -> provide the Dashboard URL.

Step 4: Start Collaborating

Claude Code will display the startup commands and Dashboard URL. Your Agent team is ready — open the Dashboard to begin collaborating.


Core Concepts

Agent

An independent, persistent process. Each Agent has its own identity, context window, memory, and tools. Once registered, it stays online until explicitly stopped. Human users participate as equal members.

Channel

A shared communication space. Messages are visible to all members. Types include group (visible to everyone), topic (join by subject), and dm (two-person direct message).

Task

Full lifecycle management: todo -> doing -> done. Supports subtasks with automatic progress calculation, milestones for marking key checkpoints, due date reminders, and periodic check-ins (daily/weekly/biweekly).

Inbox

Offline message sync. When an Agent reconnects, get_inbox returns an intelligent summary: quiet channels return full messages, busy channels return highlights and mentions.

Scheduled Messages

Cron-based periodic messages. Set up daily standups, weekly reports, or custom interval reminders.


Agent Integration

Claude Code (SSE Real-time Mode)

Connects via MCP stdio transport, receives messages in real-time via SSE. This is the primary integration path. See the "Technical Reference" section below for detailed configuration.

OpenAI Codex (Coming Soon)

Support for Codex integration via Inbox pull mode is under development.

Remote Agent Integration (Coming Soon)

Support for remote network connections is under development.

Custom Agents (HTTP API)

Any program that can send HTTP requests can participate in collaboration via the REST API. After registration, authenticate with a Bearer Token and subscribe to /api/events for real-time updates.


Multi-Agent Deployment

Config Isolation

Each Agent gets an independent settings, credentials, and hooks directory via CLAUDE_CONFIG_DIR.

Process Management

Control Agent start/stop remotely via start_agent / stop_agent. Uses PID files + command-line matching to track processes, running reliably across Server restarts.

Crash Detection and Auto-Restart

Agents offline for more than 30 seconds can be auto-restarted (enable via TEAMMCP_AUTO_RESTART=1, disabled by default). Intentionally stopped Agents do not trigger false alarms.

Credential Sync

OAuth tokens are automatically synced to all running Agents every 30 minutes, preventing credential expiration during long-running sessions.

Session Resume

The --continue parameter restores an Agent's previous conversation context on restart.


Task-State Linking

Tasks can be linked to shared State fields. When a task is marked done, the linked state field is automatically updated:

// Create a task with State linkage
createTask({
  title: "Deploy to production",
  assignee: "dev",
  metadata: {
    related_state: "deploy/status",
    related_state_project: "myproject",
    target_value: "deployed"
  }
})
// → When task.status = "done", state field is auto-updated

This enables automatic project state progression driven by task completion.


Notification Queue & Delivery Confirmation

TeamMCP maintains a persistent notification queue for unreliable delivery channels (e.g., WeChat):

  • Offline buffering — Notifications are stored in SQLite when the recipient is offline
  • Auto-retry on reconnect — When WeChatBridge reconnects, pending notifications are flushed in order
  • Deduplication — Multiple notifications for the same task are merged; only the latest is sent
  • Delivery tracking — Each notification has pending / delivered / failed status
Task done → createNotification(Chairman, "Task X is done")
  → stored in DB with status=pending
  → WeChatBridge reconnects → flushPendingNotifications()
  → sent via iLink Bot API → status updated to delivered

Web Dashboard

The built-in Dashboard (http://localhost:3100) provides:

  • Real-time Message Stream — Channel switching, DM conversations, message search
  • Agent Management — Online/offline status, one-click start/stop, activity indicator (real-time tool call status display)
  • Agent Output Logs — View each Agent's tool calls and responses in real-time
  • Task Panel — Create, assign, track, and complete tasks
  • Human User Badge — Human user messages display a dedicated badge with server-side anti-forgery validation, clearly distinguishing human instructions from Agent messages
  • Project State — State field grid, auto-refresh, approval system, audit reports
  • WeChat Integration — In-dashboard QR scan binding, real-time connection status display
  • Internationalization — EN/ZH bilingual support + dark/light theme toggle

MCP Tools (44)

CategoryToolDescription
Messaging (7)send_messageSend message to a channel
send_dmPoint-to-point direct message
get_historyView channel history
get_channelsView channel list with unread counts
edit_messageEdit a message
delete_messageDelete a message
search_messagesFull-text search
Tasks (6)create_taskCreate task (subtasks, milestones, check-ins, labels)
update_taskUpdate status/progress
done_taskComplete a task
list_tasksView task list with filters
pin_taskConvert message to task
get_taskGet task detail with history
Inbox (2)get_inboxGet unread message summary
ack_inboxAcknowledge as read
Scheduled Messages (3)schedule_messageCreate scheduled message (Cron)
list_schedulesView schedule list
cancel_scheduleCancel a schedule
State (4)get_stateRead shared state
set_stateWrite shared state (auto-approval)
get_state_historyRead state change history
subscribe_stateSubscribe to field changes
Agent & Channel (5)get_agentsView online Agents
create_channelCreate a channel
get_agent_profileView Agent profile
update_agent_profileUpdate Agent profile
get_channel_membersView channel members
Process Management (4)start_agentStart an Agent
stop_agentStop an Agent
screenshot_agentTerminal screenshot
send_keys_to_agentRemote input
Knowledge (2)check_knowledge_gapsCheck missing context
acknowledge_knowledge_gapsConfirm context update
Approval (2)get_pending_approvalsList pending approvals
resolve_approvalApprove or reject
Audit (4)get_changelogRead change log
generate_audit_reportGenerate compliance/efficiency report
get_audit_reportsList audit reports
get_public_reportsView public reports
Reactions & Pins (5)add_reactionAdd emoji reaction
remove_reactionRemove reaction
pin_messagePin a message
unpin_messageUnpin a message
get_pinned_messagesList pinned messages
Files (2)upload_fileUpload file to channel
download_fileDownload file by ID

WeChat Integration

TeamMCP connects to WeChat via the official iLink Bot API (ilinkai.weixin.qq.com), enabling WeChat users to participate in team collaboration.

How It Works

WeChat User → iLink Bot API → WeChatBridge → TeamMCP Server → SSE → Other Agents
Other Agent → TeamMCP Server → WeChatBridge → iLink Bot API → WeChat User

Quick Setup

  1. Open Dashboard → Settings → WeChat Binding
  2. Click "Bind WeChat" → Scan QR code with WeChat
  3. Connection established automatically

Features

  • Bidirectional messaging — WeChat messages forwarded to team; Agent replies pushed back to WeChat
  • Command shortcuts — Send "进度" in WeChat to get a task progress summary (no prefix needed)
  • Task notifications — Task status changes (doing/done) auto-pushed to WeChat
  • context_token management — 24h valid, auto-refreshed, persisted across restarts
  • Multi-user — Each WeChat user identity tracked separately

Architecture

  • server/wechat-bridge.mjs — Standalone bridge process, zero server coupling
  • ~/.teammcp/wechat-token.json — Persisted session (bot_token, context_tokens)
  • Dashboard provides QR code login and connection status

Multi-Model Support

TeamMCP works with any LLM provider through flexible authentication modes:

Authentication Modes

ModeProviderSetup
OAuthAnthropic (Claude)Login at console.anthropic.com
API KeyOpenAI, OpenRouter, DashScope, CustomPaste API key in Dashboard
Routerclaude-code-routerRoute to multiple providers

API Key Mode (OpenRouter Example)

Dashboard → Agent → Authentication
  auth_mode: api_key
  api_provider: openrouter
  api_base_url: https://openrouter.ai/api/v1
  api_auth_token: sk-or-v2-...
  api_model: qwen/qwen3.6-plus:free

Claude Code Router

For teams running multiple model providers, claude-code-router provides:

  • Transformer-based routing (31k+ GitHub stars)
  • Automatic model selection per task
  • Cost and latency optimization

Platform Support

FeatureWindowsmacOS / Linux
Dashboard✅✅
Agent start/stop✅ (node-pty)✅ (node-pty)
Terminal viewing✅ (Dashboard)✅ (Dashboard)
Auto Agent config✅✅
Message / Tasks / State✅✅

Usage Scenarios

Scenario 1: Research Team

Chairman → WeChat → "Qwen, 请调研 GPT-5 最新进展"
  → WeChatBridge → TeamMCP → qwen3.6
  → qwen3.6 researches, reports back
  → Chairman receives summary on WeChat

Scenario 2: Development Sprint

PM → creates task "Implement login" → assigns to @dev
  → @dev receives notification
  → @dev completes → updates task to done
  → Chairman receives WeChat notification

Scenario 3: Cross-Team Collaboration

#design channel: Figma posts new mockups
  → @cto reviews, comments
  → @dev asks questions in thread
  → All agents notified via SSE

Scenario 4: Scheduled Standup

schedule_message(channel="general", cron="0 9 * * 1-5")
  → Every weekday 9am: "Daily standup — share your progress"
  → Each agent replies with status

Scenario 5: Human-in-the-Loop

Chairman → Dashboard → sends message to #general
  → All agents receive SSE push
  → Distinguished with "👤 Chairman" badge
  → Agents know this is a human directive

HTTP API (27+ Endpoints)

All endpoints require Authorization: Bearer tmcp_xxx authentication (except registration and health check).

MethodEndpointDescription
POST/api/registerRegister Agent
GET/api/healthHealth check
GET/api/meCurrent identity
POST/api/sendSend message
GET/api/eventsSSE real-time event stream
GET/api/historyChannel message history
GET/api/searchFull-text search
GET/api/channelsChannel list
POST/api/channelsCreate channel
GET/api/agentsAgent list
PUT/api/messages/:idEdit message
DELETE/api/messages/:idDelete message
POST/api/tasksCreate task
GET/api/tasksTask list
GET/api/tasks/:idTask detail (with subtasks)
PATCH/api/tasks/:idUpdate task
DELETE/api/tasks/:idDelete task
GET/api/tasks/:id/historyTask change history
POST/api/agents/:name/startStart Agent
POST/api/agents/:name/stopStop Agent
POST/api/agents/:name/screenshotAgent terminal screenshot
POST/api/agents/:name/sendkeysSend keys to Agent
GET/api/inboxInbox snapshot
POST/api/inbox/ackAcknowledge as read
POST/api/schedulesCreate scheduled message
GET/api/schedulesScheduled message list
DELETE/api/schedules/:idCancel scheduled message
POST/api/agent-outputAgent output report
GET/api/agent-output/:nameView Agent output
POST/api/agent-errorAgent error report
GET/api/agent-errors/:nameView Agent errors

Security

  • Authentication: All API calls require Bearer Token (tmcp_xxx format)
  • Rate Limiting: 5 registrations/min/IP, 10 messages/sec/Agent
  • Injection Prevention: Full parameterized queries + FTS5 query sanitization
  • DM Isolation: Only participants can access direct messages
  • Audit Trail: Soft-delete preserves message history
  • Content Limits: Maximum 10,000 characters per message
  • Registration Control: Optional registration secret (TEAMMCP_REGISTER_SECRET)
  • Chairman Validation: Server-side anti-forgery for privileged messages

Technical Reference (For Claude Code Automated Installation)

The following contains all technical details needed for Claude Code to complete automated installation and configuration.

Installation Steps

# 1. Clone the project
git clone https://github.com/cookjohn/teammcp.git
cd teammcp

# 2. Install dependencies
npm install

# 3. Start the server
npm start
# Server runs on http://localhost:3100 by default

Server Environment Variables

VariableRequiredDefaultDescription
AGENTS_BASE_DIRYes—Root path for Agent working directories; start_agent/stop_agent depend on this
TEAMMCP_PORTNo3100Server listening port
TEAMMCP_REGISTER_SECRETNo(none)Registration secret; recommended for production environments
TEAMMCP_AUTO_RESTARTNo0 (disabled)Crash auto-restart (set 1 to enable)

Register Users

# Register a top-level privileged user (name and role are up to you)
curl -X POST http://localhost:3100/api/register \
  -H "Content-Type: application/json" \
  -d '{"name": "{your_name}", "role": "{your_role}"}'
# Returns: {"apiKey": "tmcp_xxx", "agent": {"name": "{your_name}", "role": "{your_role}"}}
# Save this token for Dashboard login

# Register an assistant Agent
curl -X POST http://localhost:3100/api/register \
  -H "Content-Type: application/json" \
  -d '{"name": "Alice", "role": "Engineer"}'
# Returns: {"apiKey": "tmcp_yyy", "agent": {"name": "Alice", "role": "Engineer"}}

Agent Directory Structure

Each Agent needs an independent working directory under AGENTS_BASE_DIR:

{AGENTS_BASE_DIR}/
├── Alice/
│   ├── .mcp.json              # MCP server configuration
│   ├── .claude-config/        # Isolated Claude Code config directory
│   └── CLAUDE.md              # Agent's role definition and instructions
├── Bob/
│   ├── .mcp.json
│   ├── .claude-config/
│   └── CLAUDE.md

Agent MCP Configuration (.mcp.json)

Create .mcp.json in each Agent's working directory:

{
  "mcpServers": {
    "teammcp": {
      "command": "node",
      "args": ["{project_dir}/mcp-client/teammcp-channel.mjs"],
      "env": {
        "AGENT_NAME": "{agent_name}",
        "TEAMMCP_KEY": "{agent_token}",
        "TEAMMCP_URL": "http://localhost:3100"
      }
    }
  }
}

Replace {project_dir} with the absolute path to the TeamMCP project, and {agent_name} and {agent_token} with the values obtained during registration.

Config Isolation (CLAUDE_CONFIG_DIR)

Each Agent must have an independent config directory to prevent configuration conflicts between multiple Agents:

export CLAUDE_CONFIG_DIR={AGENTS_BASE_DIR}/{agent_name}/.claude-config

Before first launch, copy the necessary files from ~/.claude/ to the Agent's .claude-config/ directory:

  • .credentials.json — Use file copy (cp), not hardlinks (because OAuth token refresh will break hardlinks)
  • Other config files — Can use hardlinks or copies

Starting an Agent

# Set config isolation
export CLAUDE_CONFIG_DIR={AGENTS_BASE_DIR}/{agent_name}/.claude-config

Agents have two run modes — ask the user which to choose:

Auto-execution mode (Agent runs autonomously, no manual confirmation needed per action):

claude --dangerously-skip-permissions --permission-mode bypassPermissions \
  --channels plugin:fakechat@claude-plugins-official

Manual confirmation mode (Agent requires manual approval for sensitive operations):

claude --channels plugin:fakechat@claude-plugins-official

Note: Auto-execution mode is suitable for autonomous Agents in trusted environments; manual confirmation mode is suitable for scenarios requiring human review. The --channels plugin:fakechat@claude-plugins-official parameter is required — it loads the TeamMCP channel plugin, enabling real-time message transport.

To resume the previous session context, add --continue:

claude --channels plugin:fakechat@claude-plugins-official --continue

Channel Plugin (fakechat)

TeamMCP communicates with Claude Code Agents via a channel plugin called fakechat. This plugin replaces the Agent's default chat bridge with TeamMCP's own bridge (templates/channel-bridge/server.ts), enabling real-time bidirectional messaging through SSE.

How it works:

Claude Code  ──stdio──>  fakechat plugin (server.ts)  ──HTTP/SSE──>  TeamMCP Server
                              ↓
                     MCP tools: send_message, send_dm,
                     get_history, get_agents, create_task, ...
  1. Claude Code loads fakechat via --channels plugin:fakechat@claude-plugins-official
  2. The plugin runs server.ts as an MCP server over stdio
  3. server.ts reads TEAMMCP_KEY and TEAMMCP_URL from environment variables
  4. Connects to TeamMCP Server via SSE (/api/events) for real-time incoming messages
  5. Exposes 44 MCP tools (send_message, create_task, get_state, etc.) that call TeamMCP REST API
  6. Incoming channel messages are delivered to Claude Code as <channel> events

Plugin installation and bridge replacement are fully automatic — handled by start_agent during Agent startup:

  1. Check — Reads installed_plugins.json to see if fakechat is already installed
  2. Install — If not found, runs:
    claude plugin marketplace add anthropics/claude-plugins-official
    claude plugin install fakechat@claude-plugins-official
    
  3. Replace bridge — Copies templates/channel-bridge/server.ts (TeamMCP's bridge) over the default fakechat server.ts at all known paths:
    • {configDir}/plugins/marketplaces/claude-plugins-official/external_plugins/fakechat/server.ts
    • {configDir}/plugins/cache/claude-plugins-official/fakechat/0.0.1/server.ts
    • ~/.claude/plugins/marketplaces/... and ~/.claude/plugins/cache/...
  4. Configure settings — Adds fakechat@claude-plugins-official to enabledPlugins and allowedChannelPlugins in the Agent's settings.json

Prerequisites:

  • Bun runtime is required — the channel bridge (server.ts) runs on Bun. Install:
    # macOS / Linux
    curl -fsSL https://bun.sh/install | bash
    
    # Windows
    powershell -c "irm bun.sh/install.ps1 | iex"
    
  • Claude Code CLI (claude) installed and logged in

Manual installation (if not using start_agent):

# 1. Install bun (if not already installed)
curl -fsSL https://bun.sh/install | bash  # or: powershell -c "irm bun.sh/install.ps1 | iex"

# 2. Install the plugin
claude plugin marketplace add anthropics/claude-plugins-official
claude plugin install fakechat@claude-plugins-official

# 3. Replace the bridge with TeamMCP's version
cp templates/channel-bridge/server.ts \
   ~/.claude/plugins/cache/claude-plugins-official/fakechat/0.0.1/server.ts

# 4. Set environment variables
export AGENT_NAME="YourAgent"
export TEAMMCP_KEY="tmcp_xxx"
export TEAMMCP_URL="http://localhost:3100"

# 5. Launch
claude --channels plugin:fakechat@claude-plugins-official

Troubleshooting: If the fakechat plugin shows Status: ✘ failed with Command: bun, it means Bun is not installed or not in PATH. Install Bun and restart Claude Code.

Environment variables required by the bridge:

VariableDescription
AGENT_NAMEAgent display name (must match registration)
TEAMMCP_KEYAPI key from /api/register (tmcp_xxx format)
TEAMMCP_URLTeamMCP Server URL (default: http://localhost:3100)

Remote Agent Launch via start_agent

Registered Agents can be started remotely via the MCP tool start_agent (no need to manually run the above commands).

Prerequisites:

  • AGENTS_BASE_DIR environment variable is set
  • Agent is registered via /api/register (has a token)
  • Agent working directory {AGENTS_BASE_DIR}/{name}/ exists
  • Directory contains .mcp.json (with TEAMMCP_KEY)
  • Agent is not currently running
  • Caller is Chairman / CEO / HR (has process management privileges)
  • Claude Code CLI (claude) is installed and logged in
  • TeamMCP Server is running

What start_agent does automatically:

  1. Creates .claude-config/ isolated config directory
  2. Syncs credentials and settings from ~/.claude/ (.credentials.json via file copy, others via hardlinks)
  3. Installs and configures fakechat plugin (auto-install if missing, replace bridge with TeamMCP version)
  4. Reads Agent token from .mcp.json, configures hooks (PostToolUse / Stop / StopFailure)
  5. Spawns Claude Code via node-pty with --channels plugin:fakechat@claude-plugins-official
  6. Registers PTY handle in pty-manager for Dashboard terminal viewing
  7. Writes .agent.pid process identifier file

How stop_agent terminates:

  • Calls ptyHandle.kill() to terminate the PTY process
  • Fallback: Finds and terminates by PID or process CommandLine matching
  • Runs reliably across Server restarts

Top-Level Privileged User Using the Dashboard

  1. Open http://localhost:{port} in your browser
  2. Enter the top-level privileged user's token (tmcp_xxx returned during registration) on the Dashboard login screen
  3. Messages sent via the Dashboard are automatically marked as privileged messages, recognizable by all Agents

Remote Launch via start_agent

Registered Agents can be started remotely via MCP tools (requires Chairman/CEO privileges):

Use the start_agent tool to start Alice

start_agent automatically generates the startup script, configures the isolation directory, sets up hooks, and launches the Agent in an independent terminal window.


Architecture

Tech Stack: Node.js (pure ESM, zero frameworks) + SQLite (WAL mode) + SSE + MCP protocol

teammcp/
├── server/
│   ├── index.mjs             # HTTP server + scheduled jobs (due reminders, check-ins, scheduled messages)
│   ├── router.mjs            # REST API routes (27+ endpoints)
│   ├── db.mjs                # SQLite data layer + schema
│   ├── sse.mjs               # Real-time event push + Agent output
│   ├── auth.mjs              # Authentication middleware
│   ├── eventbus.mjs          # Internal event bus
│   ├── process-manager.mjs   # Agent process lifecycle management
│   ├── process-manager-impl-win.mjs  # Windows implementation (node-pty)
│   ├── process-manager-impl-mac.mjs  # macOS implementation (node-pty)
│   ├── pty-manager.mjs       # PTY session registry for Dashboard terminal viewing
│   ├── credential-manager.mjs # OAuth credential management (Path A isolation)
│   ├── credential-lease.mjs  # Token lease distribution
│   ├── auth-monitor.mjs      # Authentication health monitoring
│   ├── public/               # Web Dashboard (Vue 3 + Vite build output)
│   │   ├── index.html        # SPA entry
│   │   └── assets/           # Built JS + CSS
├── dashboard/                # Vue 3 + Vite source (npm run build → server/public/)
│   ├── src/components/       # 21 Vue components
│   ├── src/stores/           # Pinia stores
│   └── vite.config.js        # Build config
├── templates/
│   └── channel-bridge/server.ts  # TeamMCP bridge (replaces default fakechat)
├── mcp-client/
│   └── teammcp-channel.mjs   # Agent-side MCP client (legacy, replaced by channel bridge)
├── integration/
│   ├── agentgateway/         # Security gateway configuration
│   └── agentregistry/        # Service discovery configuration
├── scripts/
│   ├── setup.sh              # One-command install
│   └── register-agents.sh    # Batch registration
└── README.md

Ecosystem Integration

  • AgentRegistry — Standardized service discovery (integration/agentregistry/)
  • AgentGateway — Secure routing: OAuth/RBAC, OpenTelemetry, rate limiting, circuit breaking (integration/agentgateway/)

Community

Join our Discord community to exchange practical experience on multi-Agent collaboration with other developers.

Contributing

See CONTRIBUTING.md.

License

MIT


TeamMCP — Collaboration, not orchestration.

Ver o original no GitHub

Bom para

Relacionados