free tool · no signup · openai-compatible
AI Coding Agent Setup Generator
Pick a coding agent and a model, and get the exact, copy-paste setup that points it at your own GPU through an OpenAI-compatible endpoint. Qwen Code, Aider, Continue.dev, Cline, the OpenAI Python SDK, and raw curl — all from one base URL.
1 — pick your coding agent
2 — pick a model preset
3 — connection (optional)
export OPENAI_BASE_URL="https://wideareaai.com/api/v1" export OPENAI_API_KEY="wai_sk_..." export OPENAI_MODEL="qwen2.5-coder-7b" qwen
⚠ Deploy the model first. The model name above must match a model loaded on a node — push any Hugging Face GGUF to a node with one click from the Models page.
⚠ Raise the context window to 32k. Coding agents send large prompts; the default 4096 tokens truncates your repo. Set it on the node detail page before running the agent.
Why one endpoint works with every coding agent
Almost every modern AI coding tool speaks the same protocol: the OpenAI Chat Completions API. That means the tool only needs three things to talk to any backend — a base URL, an API key, and a model name. Point those at OpenAI and you get GPT. Point them at Wide Area Intelligence and the exact same request is routed to a llama.cpp node running on your own GPU, with cloud failover if the node is busy or offline.
Because the wire format is identical, you never have to fork or patch the agent. You override two environment variables (the base URL and the key), set the model to whatever you deployed, and the tool behaves exactly as if it were calling OpenAI — except inference happens locally, your code never leaves your network, and there is no per-token bill.
What each agent is best at
| Tool | Best for |
|---|---|
| Qwen Code | Terminal-native agent tuned for the Qwen2.5-Coder family; strong tool use and repo edits. |
| Aider | Pair-programming in the terminal with automatic git commits and whole-repo context. |
| Continue.dev | In-editor autocomplete and chat for VS Code / JetBrains, fully config-driven. |
| Cline | Autonomous VS Code agent that plans, edits files, and runs commands step by step. |
| OpenAI Python SDK | Build your own scripts and agents against the same Chat Completions API. |
| curl | Zero-dependency smoke test to confirm the endpoint, key, and model all work. |
The two settings that make or break local coding agents
Deploy the model first. The model name in your config must match a model that is actually loaded on a node. From the Models page you can push any Hugging Face GGUF to a node with one click, so the name you paste here resolves to real hardware.
Raise the context window. Coding agents send large prompts — file trees, diffs, and multi-turn history — so the default 4096-token context is far too small. On the node detail page, set the context window to at least 32768. Too small a window silently truncates the agent's view of your repo and produces confused edits.
FAQ
Do I need an OpenAI account? No. The tools just expect an OpenAI-compatible server; Wide Area Intelligence provides one at https://wideareaai.com/api/v1 with keys that look like wai_sk_....
Can I pin a request to one specific node? Yes — send the X-WAI-Node header to route to a chosen node instead of load-balancing across all of them.
What if my node is offline? Enable cloud failover with prepaid credits, and requests fall back automatically so your agent never stalls.
/// wide area ai
These numbers are theory. Your GPU is real — put it on the network.
Wide Area Intelligence turns any machine with a GPU into an OpenAI-compatible endpoint — routed, cached, and failed over automatically. Free for 2 nodes.
Start routing — free →