Nexus gathers your AI providers into one private dashboard — one shared memory across all of them, automatic failover when a provider hits its limit, and a trace that shows exactly who answered, and why.
One dashboard houses the whole conversation — whichever AI is best for the next message, it continues right where the last one stopped.
Start a conversation with one AI and finish it with another. Nexus carries the full conversation across providers — when a chat moves, the next AI is handed the story so far, not a blank page.
A provider hits its limit or goes quiet mid-chat? Nexus moves down the line and the reply still arrives — no dead ends, no copy-pasting your question into five tabs.
Each answer shows its working: which providers were tried, which one answered, how long each took, and the honest reason the others didn't. No silent swaps, ever.
Connect the AIs you already have — paste your own provider keys, or sign in with ChatGPT, Claude and Gemini. Everything is stored encrypted; chat prefers your key, and nothing is ever shown to anyone else.
Nexus runs on its owner's own server, for a small circle of invited people. No public sign-ups, no ads, no tracking — your chats stay yours.
The composer wears many hats. Pick a mode and Nexus routes the job to a provider that can actually do it — and tells you plainly when one can't.
Everyday chat, writing and questions, answered by the strongest lane available.
Code-first models for building, debugging and reviewing — with the whole thread as context.
Answers grounded in the live web when a question needs today's facts, not last year's training.
Describe it and get a picture back — generated images stay in your chat when you save it.
Talk instead of typing, and have long replies read back to you.
Short clips from a prompt, through the providers that generate them.
Multi-step jobs — research, gather, then write it up — run as one task.
Can't decide? Auto reads the job and picks the lane, the model and the mode for you.
Your other projects can call Nexus through its OpenAI-compatible /v1 endpoint — one service key, and the same shared memory, failover and lanes answer behind it. Point any OpenAI client at Nexus and it just works.
Get your service keycurl https://nexus.130-210-56-74.sslip.io/v1/chat/completions \
-H "Authorization: Bearer <your service key>" \
-H "Content-Type: application/json" \
-d '{"model": "nexus-auto",
"messages": [{"role": "user",
"content": "Hello"}]}'
The short, honest answers — the same ones the app itself gives.
Yes. Nexus runs on free tiers and its owner's own server — there is no subscription and nothing to buy. When a provider's free limit is reached, failover simply moves your chat to another lane.
Nexus is invite-only: the owner invites a small circle of people, and there is no public sign-up — by design. If you were invited, use Sign in above.
The ones Nexus routes to — ChatGPT, Claude and Gemini, plus a bench of providers like Groq, NVIDIA, Mistral, Cohere, OpenRouter and Pollinations. Connect your own accounts and keys on the Connections page, and chat prefers them.
On the owner's server, encrypted. Saved keys are never displayed back after saving, and your chats belong to you — other members of the circle cannot see them.
One sign-in, and every provider you use is sitting on one desk.
Open Nexus