Use your local AI models in a shared team workspace
Connect a model running on your computer or server, in LM Studio, Ollama, llama.cpp or vLLM, to Nexa Agora through Nexa Bridge. Your team uses it like any other model the workspace offers, for AI colleagues and shared work.
The model runs on your hardware. Nexa Bridge calls out to the workspace over one encrypted line, so no port opens on your network.
From your computer to your team's work
Install Nexa Bridge on the computer that runs your model and connect it. The app shows a code; an owner or admin checks that it matches and approves the computer for the workspace. A member sends the approval link to them. You then pick the loaded model and the Agora model profile it serves.
1. A model is loaded in your model server, on your computer.
Give the model a place in your team's work
A model that answers on your machine becomes one of the models your workspace uses in conversations, projects and automations. The work around it stays in Nexa Agora: the colleagues involved, the documents they can use and the project your team is working on.
Owners and admins set the order of models the agents try, and can give one agent its own order. A bridged model takes its place there like any other. See how the order works.
Open source, so you can read what runs on your machine
Nexa Bridge runs on your computer, and its code is public: you can read what it does and how it calls out before you install it. Builds for macOS and Windows are signed; Linux files are verified by their build attestation. Only builds named Nexa Bridge from Nexa's repository are Nexa's. github.com/NexaEthos/nexa-bridge
What Nexa sees, along the line
The model runs on your hardware, and the app's keys stay in your system keychain. Nexa's relay carries your prompts and the answers between the workspace and your computer: the line is encrypted, but it is not end-to-end encrypted through the relay.
A bridged model doesn't make Nexa Agora an offline system: the rest of the workspace still runs on Nexa. If your organization needs a different arrangement, talk to us about a custom deployment.
No Nexa credits, a daily allowance
A bridged model spends no Nexa credits; your hardware and its running are yours. Nexa Bridge and your own provider keys share a daily token allowance set by your plan, which resets at 00:00 UTC. The workspace shows what is left today. Read about allowances and billing.
Questions before you connect
Can we use Ollama or LM Studio?
Yes, and llama.cpp or vLLM too. Load a model, then pick an Agora model profile among the models Nexa can run, marked Certified or Being tested. How well a model does depends on the model and on the work you give it.
Do we need to expose the model server to the internet?
No. Nexa Bridge calls out from the computer that runs the model, over one encrypted line, so no port opens on your network.
Does a local model mean our prompts never pass through Nexa?
No. The model runs on your hardware, but Nexa's relay carries your prompts and the answers between the workspace and your computer. The line is encrypted; it is not end-to-end encrypted through the relay.
What happens if the computer is off?
Its model can't answer until the computer, the model server and the model are back. Computers added to the same connection serve the same model and share the work.
Bring the hardware you already use
Start with one loaded model and one piece of work your team knows well: join for free, then follow the setup guide.