AgentTeams

[English](./README.md) | [中文](./README.zh-CN.md) | [日本語](./README.ja-JP.md)

**AgentTeams is an open-source collaborative multi-agent runtime platform. It enables multiple Agents to collaborate in a controlled and auditable room, with full human visibility and intervention capabilities throughout the process.** Built on a **Manager-Workers architecture**, AgentTeams features a Manager that centrally orchestrates multiple Workers, focusing on collaboration scenarios between humans and Agents, as well as among Agents within enterprise environments. AgentTeams does not compete with other Agent runtimes. Instead of implementing Agent logic itself, it orchestrates and manages multiple Agent containers (including the Manager and numerous Workers). ## Key Features - **Manager-Workers Architecture**: Eliminates the need for human oversight of individual Worker Claws by enabling Agents to manage other Agents. - **Multi-Runtime Collaboration**: OpenClaw, QwenPaw, Hermes, and experimental DeepSeek Harness Workers coexist in the same IM room. Use the runtime that best fits each role while keeping collaboration visible in Matrix. - **MinIO Shared File System**: Introduces a shared file system for inter-Agent information exchange, significantly reducing token consumption in multi-Agent collaboration scenarios. - **Higress AI Gateway**: Centralizes traffic management and mitigates credential-related risks, alleviating user concerns about security vulnerabilities in the native Lobster framework. - ☎️ **Element IM Client + Tuwunel IM Server (both Matrix protocol-based)**: Eliminating DingTalk/Lark integration overhead and enterprise approval workflows. Enables rapid user onboarding to experience the "delight" of model services within an IM environment, while maintaining compatibility with native OpenClaw IM integration. - **Integrate** [AgentLoop](https://www.aliyun.com/product/agentloop?spm=at.readme.0.0.0) ：Provides capabilities such as full-stack Agent observability and auditing, Agent evaluation and experimentation, and Agent asset management and continuous optimization. ## News - **2026-08-22**: [Release Notes](https://github.com/agentscope-ai/AgentTeams/releases/tag/v1.2.3) — AgentTeams v1.2.3: makes long-running Project workflows visible and controllable through new Controller APIs, `agt` commands, and Dashboard v1.2.4; adds audited human interventions, Project history, and QwenPaw Worker checkpoints; promotes QwenPaw as the recommended local Manager and Worker runtime; and improves Skill recovery, Windows AppService startup, and runtime reliability. - **2026-08-08**: [Release Notes](https://github.com/agentscope-ai/AgentTeams/releases/tag/v1.2.2) — AgentTeams v1.2.2: adds Manager-to-Worker custom Skill delivery with validation, storage upload, `Worker.spec.skills` assignment, and QwenPaw hot refresh and enablement without a restart; hardens Manager and Worker Skill loading; and ensures Team Leaders and Workers explicitly join Team Rooms after invitation. - **2026-08-06**: [Release Notes](https://github.com/agentscope-ai/AgentTeams/releases/tag/v1.2.1) — AgentTeams v1.2.1: unifies the Manager and Worker runtime stack on QwenPaw 2.0.1, publishes the QwenPaw Worker multi-architecture image, strengthens atomic Matrix and TeamHarness task assignment and CoPaw-to-QwenPaw state migration, adds custom model vision and reasoning capability overrides, and improves Worker lifecycle and runtime management reliability. - **2026-07-30**: [Release Notes](https://github.com/agentscope-ai/AgentTeams/releases/tag/v1.2.0) — AgentTeams v1.2.0 (stable): establishes AgentTeams naming and the final Team/Worker resource contracts end to end; adds the optional AgentTeams Dashboard; and improves Worker storage sync, Team routing and lifecycle convergence, installer support for deploying v1.1.2 with its legacy environment and storage contract (earlier releases still require the matching legacy installer), Dashboard reliability, and tooling and diagnostic safety. - **2026-07-17**: [Release Notes](https://github.com/agentscope-ai/AgentTeams/releases/tag/v1.2.0-beta.1) — AgentTeams v1.2.0-beta.1 (prerelease): completes the public rename from the retired predecessor across images, Kubernetes APIs, Helm, Matrix, storage, and runtime contracts; adds the plugin platform, TeamHarness and WorkerFlow integrations, Matrix AppService and Human SSO, model-provider routing and LLM preflight, plus richer controller observability. Beta installation requires explicit opt-in, while the stable default remains v1.1.2. - **2026-05-27**: [Release Notes](https://github.com/agentscope-ai/AgentTeams/releases/tag/v1.1.2) — AgentTeams v1.1.2: QwenPaw-first installer with keep-all upgrade flow, Team human coordinators and refreshed Team Leader coordination tools, Nacos remote skills with `sts-agentteams` / `ai-registry` STS scope, Worker CR-name decoupled from runtime name, controller reconcile metrics and graceful shutdown. - **2026-05-07**: [Release Notes](https://github.com/agentscope-ai/AgentTeams/releases/tag/v1.1.1) | [Changelog](changelog/v1.1.1.md) — AgentTeams v1.1.1: declarative MCP on Worker/Manager/Team CRDs (breaking) and on Team Leader, custom `spec.env` for CRs, Token Plan + Qwen Cloud international + `qwen3.6-plus`, namespace-scoped controller RBAC, optional `SOUL.md` in Worker packages. - **2026-04-24**: [English](blog/agentteams-1.1.0-release.md) | [中文](blog/zh-cn/agentteams-1.1.0-release.md) — AgentTeams v1.1.0: Kubernetes-native control plane, Hermes autonomous coding agent runtime, 1.7 GB image shrink, agt CLI replaces shell scripts. - **2026-04-14**: [English](blog/agentteams-k8s-native-multi-agent-collaboration.md) | [中文](blog/zh-cn/agentteams-k8s-native-multi-agent-collaboration.zh-CN.md) — Deep dive: AgentTeams as a Kubernetes-native multi-agent collaboration orchestration system. - **2026-04-03**: [English](docs/usage/resource-management.md) | [中文](docs/zh-cn/usage/resource-management.md) — AgentTeams 1.0.9: Kubernetes-style declarative resource management (YAML for Worker, Team, Human); Worker Template Marketplace; Manager QwenPaw runtime; Nacos Skills Registry and more. - **2026-03-14**: [English](blog/agentteams-1.0.6-release.md) | [中文](blog/zh-cn/agentteams-1.0.6-release.md) — AgentTeams 1.0.6: enterprise-grade MCP Server management, zero credential exposure. - **2026-03-10**: [English](blog/agentteams-1.0.4-release.md) | [中文](blog/zh-cn/agentteams-1.0.4-release.md) — AgentTeams 1.0.4: QwenPaw (formerly CoPaw) Worker support, 80% less memory. - **2026-03-04**: [English](blog/agentteams-announcement.md) | [中文](blog/zh-cn/agentteams-announcement.md) — AgentTeams open sourced under its former name. ## Why AgentTeams - **Enterprise-Grade Security**: Worker Agents operate with consumer tokens only. Real credentials (API keys, GitHub PATs) stay in the gateway — Workers can't see them, and neither can attackers. - **Fully Private**: Matrix is a decentralized, open protocol. Host it yourself, federate with others if you want. No vendor lock-in, no data harvesting. - **Human-in-the-Loop by Default**: Every Matrix room includes you, the Manager, and the relevant Workers. Watch everything. Jump in anytime. No black boxes. - **Zero Configuration IM**: Built-in Matrix server means no bot applications, no API approvals, no waiting. Just open Element Web and start chatting. - **One Command Setup**: `curl | bash` and you're done — AI gateway, Matrix server, file storage, web client, and Manager Agent. - **Skills Ecosystem**: Workers pull from [skills.sh](https://skills.sh) (80,000+ community skills) on demand. Safe because Workers can't access real credentials. ## Quick Start **Prerequisites**: Docker Desktop (Windows/macOS) or Docker Engine (Linux). **Resources**: 2 CPU cores + 4 GB RAM minimum. For multiple Workers, 4 cores + 8 GB recommended. ### Install **macOS / Linux:** ```bash bash <(curl -sSL https://raw.githubusercontent.com/agentscope-ai/AgentTeams/main/install/agentteams-install.sh) ``` **Windows (PowerShell 7+ recommended):** ```powershell Set-ExecutionPolicy Bypass -Scope Process -Force; $wc=New-Object Net.WebClient; $wc.Encoding=[Text.Encoding]::UTF8; iex $wc.DownloadString('https://raw.githubusercontent.com/agentscope-ai/AgentTeams/main/install/agentteams-install.ps1') ``` The installer walks you through: 1. Choose your LLM provider (OpenAI-compatible APIs supported) 2. Enter your API key 3. Select network mode (local-only or external access) 4. Wait for setup to complete ### Access Open http://127.0.0.1:18088 in your browser and log in to Element Web. The Manager will greet you and explain how to create your first Worker. **Mobile**: Use any Matrix client (Element, FluffyChat) and connect to your server address. **That's it.** No bot applications. No external services. Your AI team runs entirely on your machine. ## Upgrade ```bash # Upgrade to latest (preserves all data) bash <(curl -sSL https://raw.githubusercontent.com/agentscope-ai/AgentTeams/main/install/agentteams-install.sh) # Upgrade to specific version AGENTTEAMS_VERSION=v1.2.2 bash <(curl -sSL https://raw.githubusercontent.com/agentscope-ai/AgentTeams/main/install/agentteams-install.sh) ``` ## Uninstall **macOS / Linux:** ```bash bash <(curl -fsSL https://raw.githubusercontent.com/agentscope-ai/AgentTeams/main/install/agentteams-install.sh) uninstall ``` **Windows (PowerShell):** ```powershell Set-ExecutionPolicy Bypass -Scope Process -Force; $wc=New-Object Net.WebClient; $wc.Encoding=[Text.Encoding]::UTF8; $s=$wc.DownloadString('https://raw.githubusercontent.com/agentscope-ai/AgentTeams/main/install/agentteams-install.ps1'); & ([scriptblock]::Create($s)) uninstall ``` This removes all AgentTeams containers (Manager, Workers, docker-proxy), Docker volume, network, env file, workspace directory, and install log. ## Install on Kubernetes (Helm) For shared / production deployments you can install AgentTeams on any Kubernetes cluster via the official Helm chart. The default profile bundles the Higress AI gateway, Tuwunel (Matrix), MinIO and the AgentTeams controller — no external dependencies required. The commands below provide a quick installation path. See the [Kubernetes Deployment Guide](docs/usage/deployment/kubernetes.md) for cluster planning, values files, model services, runtimes, persistence, Ingress, and operations. **Prerequisites** - Kubernetes 1.24+ (kind / minikube / k3s / managed K8s — all work) - Helm 3.7+ - A default StorageClass (for the Tuwunel + MinIO PVCs) **Install (OpenAI / OpenAI-compatible)** ```bash helm repo add higress.io https://higress.io/helm-charts helm repo update helm install agentteams higress.io/agentteams \ -n agentteams-system --create-namespace \ --render-subchart-notes \ --set credentials.llmApiKey= \ --set credentials.adminPassword= \ --set gateway.publicURL=http://localhost:18080 ``` For non-OpenAI providers that expose an OpenAI-compatible API, also set `llmBaseUrl`: ```bash helm install agentteams higress.io/agentteams \ -n agentteams-system --create-namespace \ --render-subchart-notes \ --set credentials.llmApiKey= \ --set credentials.llmBaseUrl=https://your-provider.example.com/v1 \ --set credentials.defaultModel=your-model-name \ --set credentials.adminPassword= \ --set gateway.publicURL=http://localhost:18080 ``` Using Qwen (通义千问) instead ```bash helm install agentteams higress.io/agentteams \ -n agentteams-system --create-namespace \ --render-subchart-notes \ --set credentials.llmApiKey= \ --set credentials.llmProvider=qwen \ --set credentials.defaultModel=qwen3.5-plus \ --set credentials.adminPassword=