Topic LLM Backends
Proxies for Cohere: Why the Python SDK Ignores HTTPS_PROXY, and Where a Proxy Goes Instead
Where a proxy goes for Cohere, from its SDK source and docs: why HTTPS_PROXY stopped working in the Python SDK, the fix, the TypeScript SDK and base URLs.
Proxies for Meta AI: The App, Meta Model API for Developers, and the Countries Meta Allows
Where a proxy goes for Meta AI, from Meta docs and policies: the app, Meta Model API for OpenAI clients and Claude Code, team limits, and the country rules.
Proxies for Mistral: Vibe Code, the API SDKs, and the Proxy URL That Chat Apps Ask For
Where a proxy goes for Mistral, from its docs and source: /proxy-setup in Vibe Code, the Python and TypeScript SDKs, local models, regions and the terms.
Proxies for OpenRouter: The Base URL, a Network Proxy in Front, and What the Terms Say
What an OpenRouter AI proxy is, from its docs and SDK source: the base URL, a network proxy in front of your calls, the IP allowlist, regions and the terms.
Proxies for Qwen: The API Base URL in Each Region, the DashScope SDK, and Claude Code on Qwen
Where a proxy goes for Qwen, from the Alibaba Cloud docs and SDK source: the base URL of each region, the key error no proxy fixes, the SDKs and Claude Code.
Claude as a Janitor AI Proxy: What It Costs per Message, the Two Routes, and Anthropic's Own Caveat
Claude is never free on Janitor AI, but it is cheaper than most people think: at OpenRouter's listed $2 per million input tokens for Claude Sonnet 5.5, a 20,000-token roleplay message costs about four cents. The exact values for the OpenRouter route and for Anthropic's own OpenAI-compatibility layer, Anthropic's note that the layer is for testing rather than production, the usage policy that governs what Claude will write, and why no proxy of any kind changes it.
Free Gemini Proxy for Janitor AI: The OpenAI-Compatible URL, the Free Models, and the Privacy Trade
Use Google's Gemini as a free Janitor AI proxy: the OpenAI-compatible endpoint to paste, the API key from Google AI Studio, the Gemini 3.x Flash models that Google lists as free of charge, the per-project rate limits, the 18+ and region rules, and the fact that free-tier content is used to improve Google's products.
Free OpenRouter Proxy for Janitor AI: Setup, the Free Models, and the Daily Cap
Use OpenRouter as a free Janitor AI proxy: the exact Proxy URL, where the key comes from, which :free model ids exist right now (16 on 28 September 2026), the 50-a-day cap and the one-time $10 that lifts it to 1,000, the privacy setting that causes 404s, and why a second account does nothing.
Grok as a Janitor AI Proxy: The xAI Endpoint, What a Message Costs, and the EU Delay to Read First
xAI's API is OpenAI-compatible, so Grok drops into Janitor AI's proxy field: https://api.x.ai/v1/chat/completions with a key from console.x.ai and a model such as grok-4.7. What Grok costs per message at xAI's listed prices, the OpenRouter route, the acceptable-use policy that governs the content, and the region and EU-launch caveats a proxy meets but does not dissolve.
Janitor AI Error 429: What the Limit Is, Which Backend Set It, and What Actually Lifts It
A 429 on Janitor AI is a rate limit set by the backend behind your proxy: OpenRouter's 20 a minute and 50 a day, Gemini's per-project caps, DeepSeek's per-account concurrency. What each one means, the fixes that work, and the three that never do: a second account, a second key, and a different IP.
Janitor AI Network Error and 401 Unauthorized: The Two Errors That Happen Before the Model Answers
A network error on Janitor AI means the request never reached a model; a 401 means it reached one that rejected the key. Janitor's own guide traces most network errors to a wrong URL or an unrefreshed page. The exact URLs for OpenRouter, DeepSeek, Gemini, Claude and Mistral to compare character by character, a two-line curl test, the key mistakes behind 401, and the one network-error case that is about your network.
KoboldAI as a Janitor AI Proxy: A Local Model, No Key, No Token Cost, and Why the Tunnel Is the Hard Part
KoboldCpp runs an open-weights model on your own machine and exposes an OpenAI-compatible endpoint, so Janitor AI can use it with no API key and no per-token cost. The catch is the address: localhost means the machine that sends the request, so the usual route gives your endpoint a public HTTPS address through a tunnel. How the local route works, why it is the most private option, and the honest limits: your hardware, and the tunnel.
Mistral as a Janitor AI Proxy: The Direct API, the OpenRouter Route, and the $10 of Monthly Credits on Mistral's Free Plan
Mistral's models speak the OpenAI format natively, so they drop into Janitor AI's proxy field with no compatibility layer: https://api.mistral.ai/v1/chat/completions and a key from Mistral's console. What Mistral Large and Small cost per message at listed prices, the free plan on Mistral's pricing page that lists $10 a month in API credits, the OpenRouter route with today's ids, and the limits and refusals no proxy changes.
Free DeepSeek Proxy for Janitor AI: Setup and Fixing 429 Errors
Set up DeepSeek as a Janitor AI proxy in minutes: the API key, the proxy URL and the current model name, why the free route ended, and how to fix 429 errors.
Proxies that don't die mid-job
Residential, ISP, datacenter and mobile, verified by the same engine that runs tens of millions of checks. They read as a real device and hold up under load. Pay as you go, and your balance never expires. $0.44/GB is the 2,000 GB+ rate; a single gigabyte is $0.50/GB, with no minimum order.
129M+ proxy checks run · 100+ countries · HTTP / HTTPS / SOCKS · re-checked every few minutes · no signup