A detailed, practical overview of JAND AI. Read examples, technical notes, and FAQs that explain how JAND delivers fast, private, and accurate AI assistance.
All major functionality is available at no cost. There are no feature walls that block core capabilities. Monetization is limited to lightweight ads to keep the platform sustainable while preserving user access.
Conversations are stored locally in browser storage by default. No user-identifying logs are retained on our servers. When external provider calls are used (for enrichment), only the minimal request data is forwarded and only with explicit user consent.
JAND uses an LLM-first pipeline tuned for concise and context-aware replies. The pipeline adapts to conversation context to preserve intent across follow-up prompts and maintain continuity without re-sending the entire history every request.
Automatic prompt rewriting improves clarity, reduces ambiguity, and produces higher-quality outputs for code generation, content writing, and research tasks. Enhancements are deterministic and reversible.
Optional link enrichment returns curated sources and citations with answers. The feature can be toggled per session. The enrichment layer prioritizes reputable domains and returns short summaries beside each link.
Chat history is stored locally and can be exported or cleared by the user. Export formats include JSON and plain text to facilitate backups or migration to a self-hosted instance.
The UI is optimized for low-latency interactions and accessibility. It supports keyboard navigation, screen readers, and responsive layouts for mobile and desktop.
JAND's backend is modular with a FastAPI core and pluggable providers. Developers can self-host the stack for full control over data and provider selection.
Features such as formatting, code refactor, and templates run deterministically to ensure reproducible outputs useful for testing and CI workflows.
Vanilla JS + lightweight state. Emphasis on performance and low memory footprint. Local storage holds session history and theme preferences.
FastAPI-based microservice. Pluggable provider layer abstracts model vendors, search indexes, and link enrichment services. Authentication is optional for self-hosting scenarios.
By default chats stay local. Minimal server interaction occurs only when using optional features that require external data, such as link lookups. JAND does not sell personal data. For full details see our Privacy & Data Policy.
We optimized common flows to reduce latency. Examples include caching of recent provider responses, incremental context encoding, and prompt compaction to save bandwidth and speed inference.
Generate, refactor, and explain code. JAND can produce unit tests, explain errors, and translate code between languages.
Create outlines, rewrite for tone, produce titles, and summarize long text. Use prompt enhancement to refine style automatically.
Ask for step-by-step explanations, request references with Links enabled, and get compact summaries for study notes.
Yes. Core features are free. Lightweight ads fund hosting and development. There are no hidden paywalls for essential functionality.
By default your chat history is stored locally in your browser. You can export or clear it anytime. Self-hosting provides full control over storage location.
When you enable Links, the system sends a short query to configured providers. It returns curated source links and small excerpts. You can turn this off to keep all operations local.
We push updates regularly. Check the blog for release notes and changelogs. The features page will also be revised when major capabilities land.
Planned items in priority order:
Roadmap items may change based on user feedback and usage signals.
Report bugs, request features, or discuss partnerships via our social handle.