.lic) that Cosmon signs for one specific machine. Argus verifies the signature and the machine on every start, entirely offline: no license server, no account, no check-in.
What you receive
Cosmon sends your team the installer for each operating system you’re testing on, plus a provenance file for each one (.sigstore.json) that proves where the build came from — see Release integrity. The installer is the same however the machine is licensed; there is no separate on-premise build.
- Windows
- macOS
- Linux
Install: run
Cosmon Setup <version>.exe. It installs for the current user into %LOCALAPPDATA%\Programs\Cosmon and never asks for administrator rights..lic file per machine, after you send Cosmon that machine’s fingerprint (step 2).
Install and activate
1
Install the application
Install it as described in What you receive for the machine’s operating system.
2
Send Cosmon the machine fingerprint
Open Argus. On the first screen, choose Activate a license key. In the Air-gapped section, choose Copy next to the fingerprint — a 64-character code unique to this machine.Send it to Cosmon together with the machine’s hostname, so each license file can be matched back to its machine. The fingerprint is a hash; the raw hardware identifier never leaves the machine.
3
Import the license file
When the
.lic file arrives, return to the same screen, choose Import license file… in the Air-gapped section and select it. Argus opens immediately and makes no license calls from then on.The file only works on the machine it was issued for. Moving to another machine, or reinstalling the operating system, changes the fingerprint and needs a new file.4
Connect your LLM
On the Bring your own key step of setup, or later under Settings → Model & provider:
- Pick the Provider: Anthropic, OpenAI or Google.
- Enter the API key.
- For a gateway or internal model, enter its root address in Gateway URL (or Anthropic Base URL / OpenAI Base URL in Settings). Argus adds
/v1/messages(Anthropic) or/v1/chat/completions(OpenAI) itself; OpenAI reasoning and Responses-only models use/v1/responsesinstead, so a gateway serving them must expose that path too. Gateways such as LiteLLM work with the Anthropic and OpenAI providers only. - Choose Test connection. It shows Connected with the response time once the endpoint answers.
ANTHROPIC_API_KEY, OPENAI_API_KEY or GOOGLE_API_KEY environment variable instead. Prompts, responses and files go straight to this endpoint, never through Cosmon.If the machine reaches the internet through a proxy, set it first: choose Network / proxy settings on the first screen (or Settings → Network), set Proxy mode to Manual, and enter the Proxy URL. A TLS-inspecting proxy’s root CA goes in the operating system’s certificate store — see TLS inspection.5
Check it works
Open a supported CAD or CAE application, start a new chat and send a short prompt. See Quickstart for examples.
What still reaches Cosmon
The current build contacts Cosmon for the connections below, whichever license the machine uses. The machine only talks to the hosts in the Destination column; the Processed by column names the services the Cosmon backend forwards that data to.
See Data handling for what each connection carries.
Updates
To update by hand, run the new installer over the current version; chats, settings and the license are kept. Keep the previous installer in case you need to roll back.Where Argus stores data
- Windows
- macOS
- Linux
Uninstall: remove Cosmon in Settings → Apps. To also delete chats, settings and the license, delete
%APPDATA%\nexus.