AI-native by design
AI that's part of the workspace
Not a chatbot bolted on the side. A copilot that understands your work, cites its sources, and runs on the models you choose.
2 blockers detected:
• "API design" awaiting review (3d)
• "Auth migration" unassigned
Sources: Roadmap, 12 issues
The Copilot
Available on every page, project, and database.
- On a page: summarize, improve, translate, extract action items.
- On a project: identify blockers, generate tasks, status digests.
- On a database: generate insights and summaries from your rows.
- Outputs are inserted in place — and cite the sources they used.
Multi-provider AI — your choice, your key
One workspace, many providers. No per-feature AI tax.
OpenRouter (default)
One key unlocks GPT, Claude, Gemini, Llama and more through a single API.
Bring Your Own Key
Use your own provider keys, encrypted at rest with AES-256-GCM, scoped per workspace.
Custom Endpoints
Point Nexus at any OpenAI-compatible endpoint: vLLM, Ollama, LM Studio, gateways.
Enterprise AI Gateway
Run sovereign, on-prem inference with NVIDIA NIM — data never leaves your network.
Model-agnostic
Switch models per task; never locked to one vendor.
Cost transparent
Token usage and latency tracked per request, per workspace.
Honest by default
Unbuilt AI features return clear errors — never fabricated output.