Your AI assistant for code and everything in between.
BxGPT answers programming and everyday questions, streamed live from an open-weight model running on its own server. No account. No ads. Your history stays in your browser.
Under the hood
Read from the live server configuration.
- Model
- Qwen2.5 Coder · 7B
- Runtime
- Ollama, CPU inference
- Context window
- 4,096 tokens
- Longest reply
- 1,536 tokens
- History
- Stored in your browser
- Account
- Not required
Capabilities
One assistant, two kinds of work.
Whether you’re chasing a bug or drafting an email, BxGPT keeps the conversation’s context so you can follow up naturally.
For code
Developer mode, always on
Answers render with syntax-highlighted, copyable code blocks, tables and structured explanations.
Debug errors
Paste a stack trace or failing snippet and get a likely cause with a fix.
Explain code
Walk through unfamiliar code line by line, in plain language.
Refactor
Simplify, rename, split, or modernize code while keeping behavior.
Design systems
Talk through architecture, data models and trade-offs before you build.
For everything else
Plain-language help
Ask follow-up questions, change the tone, or ask for a shorter version — the conversation keeps context.
Explain concepts
From compound interest to how vaccines work, at the depth you want.
Write & edit
Draft emails, tighten paragraphs, or adjust the tone of a message.
Summarize
Condense long notes or articles you paste into key points.
Brainstorm
Generate names, outlines, plans and alternatives to react to.
How it works
Self-hosted, end to end.
There is no third-party AI API behind BxGPT. The web app and the model run on the same private server, and the model is never exposed to the internet directly.
- 1
You ask
Your message and the recent conversation are sent from your browser over HTTPS.
- 2
BxGPT checks and queues
The server validates the request, applies fair-use limits, trims history to fit the context window and waits for a free inference slot.
- 3
The model answers locally
Ollama runs the model on the server’s CPU, reachable only from localhost.
- 4
Tokens stream back
Words appear as they are generated. You can stop at any time, which also stops the model.
Privacy
Private by architecture, not by promise.
History lives on your device
Conversations are saved in your browser’s local storage. There is no server-side database of chats.
No content in logs
Server logs record timing, token counts and errors — never the text of your messages or replies.
No trackers, no cookies
No analytics scripts, advertising pixels or tracking cookies are loaded on any page.
FAQ
Good to know.
Which model powers BxGPT?
BxGPT currently runs Qwen2.5 Coder · 7B, an open-weight model served by Ollama on the same server as the app. The model is configurable, so this may change as better small models are released.
Do I need an account?
No. BxGPT has no sign-up or login. To keep the shared server responsive, each visitor is limited in how many messages they can send in a short period.
Where are my conversations stored?
In your own browser’s local storage. The server processes each message in memory to generate a reply and does not save prompts or responses. Clearing your browser data, or using “Delete all” in settings, removes them.
Why are replies slower than large cloud assistants?
BxGPT runs on a single CPU-only server rather than a data center of GPUs. Replies stream in as they are generated, usually at around reading speed, and the first reply after a quiet period can take longer while the model loads.
Can BxGPT browse the web or run my code?
No. It answers from what the model learned during training plus what you write in the conversation. It cannot open links, search the internet, or execute code.
Can the answers be wrong?
Yes. Language models can produce confident but incorrect answers, especially about recent events, niche facts and complex calculations. Review code before running it and verify important information.
Have a question? Ask it.
Free to use, no sign-up, and your chats never leave your browser.
Start a conversation