Roadmap
Where ColibotAI is headed
Version 0.8 was a full rewrite — a hybrid AI engine, custom/local models, notes and a smart token-saving cache, all on a modern, privacy-first stack. Version 0.9 made it feel live: streaming answers, follow-up questions, whole-page summarize & translate, and an adjustable summary length. Here's everything that just shipped, and what we're building next. The extension stays free, private by design, and Made in Switzerland.
Shipped Planned / exploring
Shipped
0.9 — Live answers & whole-page AI
June 2026- Streaming results — translations, summaries and explanations type out as they're generated, on-device or with your key.
- Ask a follow-up question right on any result, keeping the original context — no re-selecting.
- Summarize or translate the whole page in one click; translation runs in place, on-device.
- Adjustable summary detail — Short, Medium or Long.
Shipped
0.8 — The full rewrite
June 2026- Rebuilt on a modern stack (WXT + TypeScript + Vite, Manifest V3) with an isolated Shadow-DOM toolbar.
- Hybrid AI engine: Chrome's on-device AI first, automatic fallback to your own key.
- Modern selection toolbar, redesigned popup, options page and notes side panel.
- Full EN/IT localization, light/dark/system theming, accessibility throughout.
Shipped
Bring your own key — and your own server
0.8- Multi-provider BYOK: OpenAI, Anthropic (Claude), Google and OpenRouter.
- Custom / Local LLM — point ColibotAI at any OpenAI-compatible endpoint (Ollama, LM Studio, llama.cpp, or a remote gateway).
- No default model: you pick the model, with per-provider price/performance hints.
- Zero-token key validation on save, plus a debug mode with raw provider errors.
Shipped
Notes & smart token-saving reuse
0.8- Save any result as a searchable note; annotate, filter, import/export JSON.
- Re-running near-identical text reuses a saved answer instead of calling the AI again — no tokens spent.
- A configurable similarity threshold controls how close a match must be.
- Quick Translate panel and Search with ChatGPT for text you type, paste or select.
Shipped
Privacy & security hardening
0.8- API key stored in local storage only (never synced), with an optional session-only mode.
- Safe markdown rendering (marked + DOMPurify); least-privilege permissions; provider hosts requested on demand.
- GDPR-complete Privacy Policy and Terms; honest on-device status probing.
- Notes capped to keep storage healthy; production CSP locking the real XSS boundary.
Planned
More languages & smarter on-device
Next- Wider on-device language coverage as Chrome's built-in models expand.
- Per-site and per-language engine defaults, remembered automatically.
- Tone controls for Summarize and Explain (adjustable length already shipped in 0.9).
Planned
Notes, leveled up
Next- Tags, folders and full-text search across everything you've saved.
- Export to Markdown and copy-as-citation, alongside the existing JSON backup.
- Optional, privacy-respecting sync you control (never on by default).
Exploring
ColibotAI PRO
Future- An optional premium tier for power users — the core stays free.
- See the PRO page for the planned perks and the waitlist.
Exploring
Beyond Chrome
Future- Edge and other Chromium browsers via the same WXT build.
- Firefox, subject to its built-in-AI and Manifest V3 support.
Roadmap items are directional, not promises — priorities shift with your feedback. Have an idea? Tell us on the Support page.
Help shape what's next
The roadmap is driven by real feedback. Request a feature, report a bug, or just tell us what would make ColibotAI better for you.