The AI answer engine and research agent for the web, your files and your inbox. Every answer cited.
Built on open models & open infrastructure
GJAM is more than a chatbot. It is an answer engine that cites, a research agent that shows its work, and a home for your own knowledge and agents.
Live web search, page reading and video results — every claim linked to its source so you can verify in one click.
Usually yes, if you drive 1,000+ km a month. Running costs are roughly ₹1–1.5/km vs ₹6–7/km for petrol1, and a higher purchase price is partly offset by subsidies2. Battery warranties of 8 years cover the riskiest component 3.
Plans the angles, reads a dozen sources in parallel and writes a cited report while you watch.
Drop in PDFs, notes and links. Ask questions across them with page-level citations.
Give an AI a role, instructions, knowledge and tools. Share it with a link.
Connect Gmail, Drive and Calendar. GJAM searches them live — nothing is copied or stored.
Type a question, mention a Space with @, or pick an agent. Choose which sources it may use.
It searches the web and your sources in parallel, reads the best pages and cross-checks facts.
Every claim links to where it came from. Follow up, go deeper, or save it to a Space.
Start from a template, tweak the instructions, attach your Spaces — or describe what you need and GJAM drafts the agent for you.
| GJAM | Typical AI chat | |
|---|---|---|
| Live web answers with clickable citations | ||
| Ask questions across your own PDFs & notes | ||
| Search your Gmail, Drive & Calendar | ||
| Build & share custom agents — free | ||
| Watch research happen step by step | ||
| Open models, no sign-up required |
Comparing tools? Read GJAM vs Perplexity.
Yes. Search, Deep Research, Spaces and Agents are free to use with no sign-up. Your data is tied privately to your browser.
From live web search and the pages GJAM reads, plus any Spaces or Google sources you switch on. Every claim is cited so you can check it.
No. GJAM requests read-only access and queries Google live when you ask. Only the encrypted access token is stored, and you can disconnect anytime.
Open models such as Llama, Gemma, DeepSeek and Mistral served via NVIDIA NIM. You can switch models from the picker in the search box.
PDFs, Word documents, text, Markdown, CSV, HTML and code files up to 10 MB each, plus web links and pasted notes.
Free. No sign-up. Cited answers from the web and your own knowledge.
