LLM API Chat Debugger

LLM API Chat Debugger helps developers, PMs, and AI users quickly test large language model APIs in the browser. Enter an OpenAI-compatible Base URL, API key, and model name to run multi-turn chat and raw API debugging—no desktop client or proxy server required.

The tool is local-first and privacy-minded: keys and chat history stay in your browser by default, and requests go directly to the endpoint you configure. This site does not proxy traffic. If the provider allows CORS, you can try models much like an official playground.

Key features

• Chat mode: multi-turn conversations, streaming replies, Markdown rendering, plus copy, edit, and regenerate.
• API Debug mode: edit the full request JSON and inspect raw responses, status codes, and latency.
• API settings: Base URL, API Key, model, Temperature, Max Tokens, and System Prompt, with common provider presets.
• Fetch models: call /models to list available models and fill one in quickly.
• Attachments: upload text files or images (for multimodal models) to add context.
• Recent chats: keep recent sessions locally, restore them, export JSON, or delete them.

Who it is for

• Verifying self-hosted or third-party OpenAI-compatible gateways
• Comparing answers from different models with the same prompt
• Tuning temperature, max tokens, and system prompts
• Quick chats against local stacks such as Ollama without building a UI
• Demoing API behavior or exporting conversations for documentation

How to use

1. Fill in Base URL (usually ending with /v1, without /chat/completions), API Key, and model. Use the help tip or provider dropdown if unsure.
2. Adjust Temperature, Max Tokens, and System Prompt as needed. Turn on Remember Config to store settings in localStorage only.
3. Send messages in Chat mode, or switch to API Debug to craft JSON and inspect raw output.
4. Attach files, stop an in-flight request, start or clear a chat, and restore history from Recent chats.

Compatibility

Requests follow OpenAI-style /chat/completions and /models paths, so most compatible cloud providers and local inference servers work. If the browser reports a CORS error, enable CORS on the API or use a gateway that allows browser access.

FAQ

Q: Does this site store my API key?

A: No server upload. With Remember Config on, the key is saved only in localStorage. On shared computers, keep it off and clear site data afterward.

Q: Why do I see a CORS failure?

A: The browser calls your Base URL directly. If the API does not allow cross-origin requests from this site, the browser blocks them. Check URL, key, network, and CORS settings.

Q: How should I write the Base URL?

A: Use the version prefix only—often ending with /v1. The tool appends /chat/completions. Do not put the full chat path in Base URL.

Q: Chat vs API Debug?

A: Chat is for multi-turn conversation and rendered replies. API Debug is for hand-editing JSON and inspecting streaming raw output and HTTP status during integration.

Q: Which providers are supported?

A: Any OpenAI-compatible endpoint can be tried. Built-in presets speed up common providers; model names follow each vendor’s docs, or use Fetch models to pick one.