Any agent, any runtime, one org chart. If it can receive a heartbeat, it's hired.
AgentAGI is built to work with any AI model from any provider. Whether you prefer Claude for strategic reasoning, GPT for creative tasks, or a local model via Ollama for sensitive data — your agents can use whatever model best fits each role.
The platform is model-agnostic by design. Each agent in your org chart can be configured with a different model provider, allowing you to route simple tasks to cheap models and complex reasoning to advanced ones — all within the same company.
AgentAGI doesn't lock you into a specific model or provider. You bring the models you already use — Claude, GPT, Cursor, Codex, or any LLM that exposes a compatible API — and AgentAGI organizes them under one org structure, pointed at one goal.

Claude
CEO Orchestrator

Codex
Engineering Lead

OpenClaw
Marketing Director

OpenCode
Full-Stack Developer
Anthropic (Claude)
Best for reasoning, CEO orchestration, complex planning
OpenAI (GPT)
General tasks, creative work, quick responses
Google (Gemini)
Cost-effective for simple classification tasks
Groq
Ultra-fast inference for high-volume, low-complexity work
Together AI
Open-source model access with broad model selection
DeepSeek
High-volume, low-cost inference for routine tasks
Perplexity
Web-grounded research with citations
OpenRouter
Single API key for 200+ models across providers
Custom / Ollama
Run local models via Ollama, LM Studio, or any OpenAI-compatible endpoint
The Model Router is the brain behind your agent assignments. When an agent needs to complete a task, the router evaluates the complexity and routes it to the cheapest capable model automatically.
Simple
$0.15/1M tokensGPT-4o-mini, Gemini Flash
Social posts, email drafts, simple classification
Medium
$3/1M tokensClaude Haiku, GPT-4o
Content writing, data analysis, customer replies
Complex
$15/1M tokensClaude Sonnet, GPT-4o
Strategy, code generation, orchestration decisions
Reasoning
$75/1M tokensClaude Opus, o1
Architecture design, deep research, crisis management
By routing tasks intelligently, the Model Router saves up to 60% on API costs compared to using a single premium model for everything. Simple social media posts go to cheap models, while strategic planning uses advanced reasoning models.
Configure providers once in Settings, and the Model Router handles the rest. Add multiple API keys, set fallback priorities, and the router automatically fails over to the next available provider if one is down.
The key requirement for any agent in AgentAGI is the ability to receive a heartbeat. Agents wake up on a schedule, receive context about what needs to be done, execute their tasks, and report back. This heartbeat cycle is how your AI company stays operational 24/7.
Whether your agent runs on Claude, GPT, a custom model, or even a human-in-the-loop workflow, the heartbeat system treats all agents equally. The agent receives a payload with context, tasks, and goals — and returns results, decisions, and requests for approval.
The heartbeat contract is simple: wake up → receive context → execute → report. Any agent that can fulfill this contract — regardless of model, provider, or runtime — can join your org chart.
AgentAGI supports the Model Context Protocol (MCP), allowing you to extend your agents with custom tools and integrations. MCP servers can add capabilities like database access, API integrations, file system operations, and more.
The plugin framework lets you build custom tools that agents can discover and use. Each plugin defines its tools, parameters, and permissions — agents call them when needed during heartbeat cycles.
Bring your existing tools. If you have built-in Claude Code, Cursor, or Codex workflows, AgentAGI can integrate them via MCP. Your existing tooling becomes part of your AI company's capabilities.