Smart Router · Integrations
Use the Smart Router with your tools
The router is an OpenAI-compatible endpoint, so any tool that lets you set a base URL works with it. Every guide below uses the same three values: the base URL, your aism_ key and a routing mode as the model. Your own provider keys, zero token markup, and 1,000 routed requests a month on the Free plan.
1,000 routed requests a month free · your own provider keys · no card
Base URL: https://aistupidlevel.info/v1 API Key: aism_your_key_here Model: auto-coding (or auto, auto-reasoning, etc.)
01
What every tool needs: a base URL, a key and a model
The same three values go wherever a tool asks for an OpenAI-compatible API. The quality modes read the same scores the leaderboard shows.
- Base URL
https://aistupidlevel.info/v1- API key
aism_your_key_here: your router key from API keys, sent asAuthorization: Bearer. If a tool insists on ansk-prefix,sk-aism_…is accepted.- Model
- A routing mode such as
auto-coding,auto-toolingorauto-best, or any real model id fromGET /v1/modelsto pin one model.
22 models are ranked on the Combined board in the last 24 hours · as of 26 Oct 2026, 10:51 UTC · the leaderboard
Claude Code is the one exception: it speaks the Anthropic Messages API, and the router passes it through natively at POST /v1/messages with ANTHROPIC_BASE_URL set to the site root. The Claude Code guide has the two variables.
02
Which tools work with the Smart Router?
One guide per tool: the exact settings, the routing modes that fit, what the tool supports behind the router, the errors it can show, and a FAQ.
Coding agents
Agents that edit repositories and call tools: auto-coding for the code, auto-tooling for long agent sessions, auto-value-coding to spend less per run.
- Claude Code
- Two environment variables send Claude Code through the router’s native Anthropic Messages passthrough: your own Anthropic key, no token markup, and the same allowance, budgets and request logs as every other endpoint.
- Cline, Kilo Code and Roo Code
- Pick the OpenAI Compatible provider in Cline, Kilo Code or Roo Code, paste the base URL and your
aism_key, and use a routing mode as the model ID: your own provider keys, no token markup, and the model that answered in every response. - Aider
- Two environment variables and a model flag put Aider on benchmark-routed models:
OPENAI_API_BASE,OPENAI_API_KEYand--model openai/auto-coding, with your own provider keys and no token markup. - OpenCode
- OpenCode takes the router as a custom OpenAI-compatible provider: one click through CC Switch, or a provider block in
opencode.json, with your own provider keys, no token markup andauto-codingas the default model.
Editors
Editor assistants that take a custom OpenAI base URL: auto-coding as the model name, with the tool’s own limits noted truthfully.
- Cursor
- Cursor’s Override OpenAI Base URL points chat and plan mode at the router: your own provider keys, no token markup, and
auto-codingas the model name. The MCP server adds rankings and route previews to the editor. - Continue
- One block in
~/.continue/config.yamlgives Continue’s chat, edit and apply roles a benchmark-routed model: your own provider keys, no token markup, and the model that answered in every response header. - Zed
- Zed’s OpenAI-compatible provider takes the router’s API URL, your
aism_key and a routing mode as the model; the key is kept in the system keychain, and the models can also go insettings.json.
Frameworks
Libraries that build agents and chains: auto-best for general work, auto-tooling for agents that call tools, auto-reasoning for analysis.
- LangChain
langchain-aistupidleveladdsChatAISL, a LangChain chat model on the Smart Router: tools, structured output and streaming, your own provider keys, no token markup, and the model that answered and why inresponse_metadata.
Automation
Workflow tools: the Chat Model sub-node for agents and chains, and benchmark data as workflow steps.
- n8n
- The
n8n-nodes-aistupidlevelcommunity node adds an AI Stupid Level node for chat, routing and benchmark data, and an AI Stupid Level Chat Model that plugs into the AI Agent, so every call, tool turns included, goes through the router.
Chat interfaces
Self-hosted chat front ends: every routing mode as a model in the menu, with the model that answered shown per reply.
- Open WebUI
- The AI Stupid Level Function on openwebui.com gives Open WebUI every routing mode as a model, with the model that answered and why above each reply; a plain OpenAI connection on
https://aistupidlevel.info/v1works too. - LibreChat
- A custom endpoint in
librechat.yamlputs the routing modes in LibreChat’s model menu, withfetch: trueto list every model the router offers andauto-fastest-qualityfor conversation titles.
03
Ready-made integrations: packages and plugins
Published packages that add AI Stupid Level as a provider, from the Integrations section of the docs. Use your router key wherever a tool asks for an API key, and a routing mode as the model.
n8n-nodes-aistupidlevel- n8n. A node for chat, routing and benchmark data, plus an AI Stupid Level Chat Model for the AI Agent. Install it from Settings → Community Nodes.
@aistupidlevel/mcp- MCP server, listed in the official MCP Registry. Rankings, drift, incidents, route previews and router answers as tools in Claude, Cursor, VS Code and any MCP client.
langchain-aistupidlevel- LangChain chat model ChatAISL: tools, structured output and streaming, plus the model that answered and why.
@aistupidlevel/ai-sdk-provider- Vercel AI SDK provider: generateText, streamText and tools.
@aistupidlevel/dsh-provider- DeepSeek Harness plugin bundle: adds AI Stupid Level as a provider route with the routing modes as models.
@aistupidlevel/openclaw-provider- OpenClaw provider plugin on ClawHub: the routing modes as OpenClaw models, with a live model list.
aistupidlevel (Hermes)- Hermes Agent provider plugin in the official Hermes plugin catalog: hermes plugins install aistupidlevel.
AI Stupid Level (Dify)- Dify model provider plugin on the Dify Marketplace, with the routing modes as models.
AI Stupid Level (RAGFlow)- RAGFlow model provider: pick it under Model providers and paste your router key.
@aistupidlevel/pi-provider- Pi provider extension: the routing modes as Pi models, with the model that answered in the footer.
lfx-aistupidlevel- Langflow model provider and component: AI Stupid Level in the Language Model and Agent pickers.
AI Stupid Level (Open WebUI)- Open WebUI Function on openwebui.com: every routing mode as a model, with the model that answered and why above each reply.
04
Questions about connecting tools to the Smart Router
Which tools work with the Smart Router?
Any tool that lets you set an OpenAI base URL: the router is an OpenAI-compatible endpoint at https://aistupidlevel.info/v1. This page has a guide for Claude Code (through the native Anthropic Messages passthrough), Cursor, Cline with Kilo Code and Roo Code, Continue, Aider, OpenCode, n8n, LangChain, Zed, Open WebUI and LibreChat. The router documentation adds samples for the OpenAI SDK, LlamaIndex, the Vercel AI SDK, LiteLLM, CrewAI, browser-use, Instructor, promptfoo, OpenClaw, Hermes Agent, DeepSeek Harness, Open Interpreter, Dify, RAGFlow, Cherry Studio, AnythingLLM, Crush, Pi, Langflow, OpenHands and Activepieces.
What do I enter as the model?
A routing mode: auto-coding for code, auto-tooling for agents that call tools, auto-reasoning for analysis, auto-best for the best combined score, and the auto-value* modes for the most points per measured dollar among near-top models. Any real model id from GET /v1/models pins that model and bypasses routing. Every mode is defined in the Virtual models section of the docs.
Does this cost extra?
No. You pay the providers directly for tokens, with your own keys; the router adds nothing to token prices. The Free plan includes 1,000 routed requests a month (Pro Intelligence 10,000, Developer 100,000, Teams 1,000,000), and prepaid credits cover any beyond that at $5 per 10,000. Only successful requests count. No card for the Free plan.
Does Cursor work with the Smart Router?
In chat/plan mode, yes: Settings → Override OpenAI Base URL with https://aistupidlevel.info/v1 and your aism_ key. Composer, inline edit and autocomplete are locked to Cursor’s backend. The Cursor guide has the steps, and the MCP server adds rankings and route previews to the editor.
Does Claude Code route between models?
No. Claude Code speaks the Anthropic Messages API, and the router passes it through natively: the request goes to Anthropic, with your Anthropic key, for the Claude model Claude Code names, with the same allowance, budgets and request logs as every other endpoint. The Claude Code guide covers the two environment variables.
Do streaming and tool calling work?
Yes. Streaming is real: text arrives as the model writes it (X-AISM-Streaming-Mode: native), tool calls arrive in one chunk each once complete, then usage if you set stream_options.include_usage. Tool calling works end to end with OpenAI-format tools; the router remembers each model’s own turn for 24 hours, so multi-step agent loops keep working, and a conversation stays on the model that started it. The router is text only: images, audio and files get a 400 unsupported_content.
One key, every tool
A free account gives you a router key, the Providers page for your own keys, and the request allowance. No card.
1,000 routed requests a month free · your own provider keys · no card