AirLLM's Air Router is an intelligent routing layer that connects your coding tools, agents, and applications to leading AI providers. Choose one AirLLM routing model, use one API key, and each request is automatically matched to the best-fit underlying model for the job — balancing quality, speed, and cost.
AirLLM provides the routing layer. The underlying AI models are provided by leading model providers.
Your tools & agents
One AirLLM endpoint
Automatic routing
Best-fit underlying model
Every AirLLM model is a routing model. Your selected alias tells AirLLM which provider family or use case you prefer, and AirLLM automatically matches each request to the best-fit underlying model.
airllm-auto
A cross-provider routing model that automatically matches every request to the best-fit coding model across supported providers. Quick edits receive fast, efficient handling, while difficult coding problems receive the capability they need — no manual selection or tuning required. Works in any compatible AI coding agent.
Best for: developers who want the best default balance of quality, speed, and cost across providers.
airllm-anthropic
A routing model that automatically matches each request to the appropriate model within Anthropic's Claude family. Use it when you prefer to keep your requests within the Anthropic provider family while still benefiting from automatic routing. It can be used in any compatible coding agent and is not limited to Claude Code.
Best for: teams and developers who prefer the Anthropic model family.
airllm-openai
A routing model that automatically matches each request to the appropriate model within OpenAI's top and latest models. Use it when you prefer to keep your requests within the OpenAI provider family while still benefiting from automatic routing. It can be used in any compatible coding agent and is not limited to Codex CLI.
Best for: teams and developers who prefer the OpenAI model family.
airllm-coding
A coding-focused routing model for questions and development tasks outside a full coding agent. Use it to explain code, diagnose errors, review snippets, generate functions, discuss architecture, or get technical help from a chat interface or API application.
Best for: coding questions in chat apps, internal developer assistants, and API-based coding help.
airllm-thinking
A reasoning-focused routing model for tasks that require deeper analysis before producing an answer. Use it for planning, research, architecture decisions, comparisons, difficult analysis, and complex general problem-solving.
Best for: deep reasoning, planning, analysis, research, and complex decisions.
airllm-agent
A routing model designed for custom agents, automation systems, and multi-step AI workflows. It is designed for consistent routing throughout long-running workflows while helping teams maintain quality and manage usage efficiently from the first step to the last.
Best for: custom agents, workflow automation, multi-step pipelines, and agentic applications.
All AirLLM routing models work in compatible AI coding agents, API applications, and custom workflows. Choose based on the provider family or use case you want — not the tool you use.
Want AirLLM to choose across supported providers?
airllm-auto
Prefer Anthropic's Claude family?
airllm-anthropic
Prefer OpenAI's model family?
airllm-openai
Asking coding questions from a chat application?
airllm-coding
Working on analysis, planning, research, or deep reasoning?
airllm-thinking
Building your own agent or automation pipeline?
airllm-agent
Still not sure? Use airllm-auto — it is the recommended default and automatically routes across supported providers.
All of these routing models can run through AirCode CLI, Claude Code, Codex CLI, Cursor, compatible Copilot-style workflows, custom agents, and other supported AI tools. Pick by provider family or use case, not by tool name.
Use AirLLM across multiple coding tools and development sessions without managing separate provider integrations.
A single AirLLM API key can power AirCode CLI, Claude Code, Codex CLI, Cursor, other compatible coding tools, and custom agents. Manage usage through one account, one bill, and one dashboard.
Any AirLLM routing model can be used in any compatible agent. Point AirCode CLI, Claude Code and Codex CLI to the same routing model, use different models for each tool, or run multiple Cursor sessions with different routing preferences. The model alias does not lock you to a specific coding tool.
Run multiple coding agents in parallel on separate tasks, worktrees, or branches:
AirLLM provides one routing layer for all active sessions.
Use airllm-auto when AirLLM should route across supported providers. Use airllm-anthropic or airllm-openai when your team has a provider-family preference.
Use a separate coding session for each distinct task. Focused sessions can reduce unnecessary context, help answer relevance, and make usage easier to understand.
Give each coding agent:
Use the AirLLM dashboard to monitor usage across tools, API keys, developers, and agent workloads.
AirLLM routing models can be used with compatible coding tools, agents, and applications that support custom AI providers or external API endpoints.
Tool compatibility depends on whether the tool supports a custom endpoint or model provider. Refer to the setup documentation for verified configuration instructions.
Choose an AirLLM routing model, connect your preferred coding tool or application, and let AirLLM automatically match every request to the right underlying model.
Keep exploring
Intelligent LLM routing for AI coding agents: the right model for each step.
Read ExplainerHow routers choose a model for each request.
Read ExplainerWhy coding agents need per-step model routing, and what to look for in a router.
Read