Local-first workspace context layer for AI.
Locus sits over your desktop so you can capture text, screenshots, notes, code snippets, terminal output, and browser context without copy-paste chaos. Stage messy context, organize it into clean thread memory, then use local or cloud models only when you choose.
Windows builds are published from the GitHub Releases page.
- Installer:
Locus Setup.exe - Portable:
Locus portable.exe
Download the latest build from:
If a release does not have assets yet, the workflow may still be building or a release tag has not been pushed.
- Capture workspace context — highlight text, grab screenshots, capture terminal output, or create notes.
- Stage before saving — use Working Context to amend, clean, and organize context before it becomes memory.
- Save to thread memory — context cards stay attached to persistent threads instead of disappearing into chat history.
- Inspect provenance — keep track of source app, source window, capture type, timestamps, and saved context.
- Use quick actions — ask, explain, summarize, simplify, fix, review, diagnose, rewrite, OCR, and more.
- Customize the radial menu — choose the six actions you want in the quick action wheel.
- Use local or cloud models — route everyday work to local models and reserve cloud models for deeper tasks.
Capture context
→ stage it in Working Context
→ edit or organize it in Note / Document / Screenshot mode
→ save it to a Locus thread
→ search, review, reuse, or send it to a model later
The goal is simple:
Locus gives AI the context you choose, organized the way you want, with receipts for where it came from.
A temporary staging area for selected text or captured material. You can add more context, ask a quick question, amend the content, or save it into a Locus thread.
Saved memory units attached to a thread. Cards can represent selected text, screenshots, notes, documents, model results, or other captured context.
A side panel for deeper work and saved context review. Current directions include context cards, source grouping, timeline view, saved items, and model/action settings.
Capture a screenshot, write or edit a note about it, then save the clean note and original screenshot provenance to Locus.
A configurable quick-action menu that appears near captured context. Pick the actions you actually use.
Locus is designed around three model tiers:
| Tier | Purpose |
|---|---|
| Quick | Local, small, private, fast. Good for tags, titles, summaries, cleanup, and simple explanations. |
| Standard | Balanced quality/cost. Good for normal asking, code explanations, and medium context. |
| Deep | Premium/high-quality. Good for architecture, hard debugging, safety review, and final checks. |
Locus should never silently move private local context to a cloud model. Cloud escalation should be explicit.
- Locus is local-first.
- Captured text is not sent to a model until you explicitly run an action.
- Local models can be used for private/quick work.
- Cloud providers are optional and should be treated as explicit routing choices.
- Audit/provenance data should track what was sent, where it went, and which model handled it.
git clone https://github.com/NovasPlace/Locus.git
cd Locus
npm ci
npm startBuild packages:
npm run build:win # Windows NSIS installer + portable exe
npm run build:linux # AppImage + deb
npm run build:mac # dmgWindows release artifacts are built by GitHub Actions when a version tag is pushed.
git tag v4.2.4
git push origin v4.2.4The release workflow builds the Windows app and uploads the installer/portable executables to GitHub Releases.
src/
├── main.js # Electron main process, IPC, windows, providers
├── index.html # Main overlay UI
├── companion-panel.html # Companion panel UI
├── quick-panel.html # Working Context / quick context window
├── platform.js # Cross-platform helpers
├── thread-store.js # Thread/context persistence
└── model/ # Model router/provider layer when enabled
Near-term focus:
- First-class thread/context/message data model
- Unified Context Panel
- Source-based context grouping
- Context timeline
- Smart tagging and filtering
- Local-first model router
- Interaction audit log
- Screenshot understanding / OCR
- Markdown/JSON export
Later:
- Knowledge graph
- Plugin/action system
- Cross-device sync
- Collaboration/sharing
MIT — © Donovan Everitts / NovasPlace