Show HN: Korvo – Local-first AI workspace that traces answers to the source
6 hours ago
- Korvo is a local-first desktop app for macOS that acts as a private research workspace, running entirely on the user's machine.
- It supports a full decision workflow: capture files, plan analysis, verify claims, challenge assumptions, and record decisions.
- Korvo indexes content locally, detects project goals, identifies missing context, and proposes analysis plans for user approval.
- All conclusions and claims are grounded with citations linking back to source documents in the workspace.
- The tool uses a dual-model critique system where one model builds a case and another critiques it, keeping revisions and unresolved questions visible.
- Privacy is emphasized: files, indices, notes, and history stay on-device, with cloud models used via user's own API key and optional offline operation via Ollama.
- Pricing includes a free Explorer tier with one project and 15 files, and a Pro subscription at $149/year (planned $299/year) for unlimited projects and full workflow.
- Korvo does not resell model tokens; cloud API usage is billed separately by the provider, with typical personal use costing $5–$30 per month.
- Integration with MCP servers is supported, allowing tools like Superhuman Mail, but models cannot autonomously execute tools without user confirmation.
- The product is in early access, with the two-model critique and Medha quality measurement layer still in beta testing.
- Exports to PDF and Markdown are available, and users can delete local workspace data at any time.