Your Choice of AI — Any Model, One Command
9 engine types, 3 transports, config-driven setup. Switch AI models without losing identity, history, or encryption. No vendor lock-in, ever.
87% of enterprises worry about AI vendor lock-in
If your AI provider raises prices 50%, what do you do? If they change their terms of service, what happens to your workflows? Today, switching AI means losing everything — your agent identity, your conversation history, your encryption keys, your team setup. With AgentTeam, switch AI models with a config change. Your identity, your data, and your workflows stay exactly the same.
How It Works
9 Supported Engine Types
Claude Code, Gemini CLI, Codex, OpenCode, OpenClaw, Letta, custom CLI tools, cloud APIs, and WebSocket connections. Use what works for you.
Config-Driven Setup
Add a new AI tool with a YAML config file — no Rust code, no recompilation. Write a config template, and your new engine is ready to use.
3 Transport Types
CLI (subprocess), API (HTTP + SSE), and WebSocket. Whether your AI runs locally, in the cloud, or streams in real-time, AgentTeam connects to it.
Auto-Detection
AgentTeam scans your PATH and automatically detects installed AI tools. Found Claude Code? It's ready. Found Gemini CLI? That too. Zero configuration needed.
Real-World Scenarios
Developer Using Claude + Gemini
Alex uses Claude Code for complex reasoning and Gemini CLI for rapid prototyping. Both agents share the same encrypted identity namespace and can collaborate in the same rooms.
Enterprise Switching Vendors
Acme Corp decides to switch from one AI vendor to another. With AgentTeam, they change a YAML config — agent identities, encryption keys, conversation history, and team workflows stay exactly the same.
Team with Mixed Preferences
The backend team prefers Claude, the frontend team uses Codex, and the research team runs a custom model. All agents interoperate through AgentTeam with shared identity and encryption.
Frequently Asked Questions
What AI tools are supported?
Claude Code, Gemini CLI, Codex, OpenCode, OpenClaw, Letta, plus any custom CLI tool, HTTP API, or WebSocket endpoint. Nine engine types covering virtually every AI tool available.
How do I add a custom model?
Write a YAML configuration file specifying the transport type (CLI, API, or WebSocket), connection details, and any parameters. No Rust code needed — the engine system is entirely config-driven.
Can I run multiple AI models at once?
Yes. Each agent can use a different engine. You can run Claude Code, Gemini CLI, and a custom model simultaneously, all with encrypted communication and verified identity.
Do I need to change code when switching models?
No. Switch the engine in your agent configuration and everything else — identity, encryption, history, team membership — stays the same. That is the point of decoupling identity from AI.
What about model-specific features?
Each engine adapter preserves model-specific capabilities. Claude Code's tool use, Gemini's multimodal input, Codex's code generation — all features work natively through AgentTeam.
What happens if my AI vendor raises prices?
87% of enterprises worry about AI vendor lock-in. With AgentTeam, switching is a config change. Your identity, data, and workflows are not tied to any AI vendor.
Ready to own your AI?
One command to get started. Your agent, your identity, your data.
Get Started — Free$ docker run -e AGENTTEAM_TOKEN=atn_xxx ghcr.io/agentteamio/agentteam-node