Import chats
Paste an AI conversation or drop in a share link. Shelvia detects the source and pulls out the parts worth keeping.
Keep project memory where the work happens.
Shelvia is more than a place to save chats. It holds your scattered AI work, points you to the right tool for the next step, and keeps the reasoning that produced each result close to the result itself.
Bring the scattered pieces in.
Paste, drop, drag, or upload. Source detection runs in the background, so a chat, screenshot, or link arrives already tagged with where it came from.
Save screenshots
Drop in a screenshot of a chat, prompt card, or research finding. Text comes out, source detection runs, and it joins the project memory.
Add links
Save the source behind a claim, not just the claim itself. Domain, type, and freshness are detected automatically.
Upload files
PDFs, exports, transcripts. Files land alongside the rest of the project so a fact and the document it came from stay together.
Manual notes
Jot a thought without a chat behind it. Notes carry the same project + tag structure as everything else.
Save prompts
When a prompt produces something worth keeping, Shelvia flags it. You confirm the save with one click.
Approve what becomes memory.
AI-generated, imported, and agent-proposed memory enters review before it becomes trusted. Workspace members approve, edit, or reject each candidate; human-approved project edits are audit logged.
Review queue
Every import, connector sync, and API write enters a queue first. Nothing reaches trusted memory before a workspace member approves it.
Status labels
Source-backed, matches your goal, possible duplicate, needs review, check source. Each label tells you what to do next without reading the whole row.
Pin decisions
Save a choice as a decision with the reasoning behind it. Decisions become a traceable record of what changed and why.
Preserve sources
The link, citation, or transcript that backed the call stays attached to the decision. Not lost in a sidebar later.
Record what changed
When a decision is superseded, the new one points at the old one. The history of reversals stays readable.
Let meaning form around the work.
Related material clusters around projects. Decisions, sources, prompts, and notes stay close to the work that produced them. Health verdicts surface stale or conflicted rows.
Living project memory
Each project keeps a current summary, key decisions, open questions, prompts that worked, and recommended next moves. Refreshed as the work evolves.
Source-aware notes
Highlight a passage from any imported chat or document and save it as a note. The original passage stays linked.
Tags and collections
Lightweight grouping for the cases where projects aren't the right shape. Research themes, prompt families, recurring questions.
Smart grouping
Related work clusters together. Duplicate ideas, follow-up threads, source mix-ups all surface where they belong.
Memory health verdicts
Stale, conflicted, low-confidence, and missing-source memory surfaces with an explicit verdict. The same verdict travels into context packs and the API.
Make trusted memory reusable.
Search, context packs, and rolling summaries surface what is worth using next. Tools and agents read the same trusted set through API, MCP, and SDK.
Unified search
One hybrid keyword + vector search across imports, prompts, sources, decisions, and notes. Find what you saved without remembering where.
Context packs
Ranked, source-backed context for the next session. Decisions, sources, prompts, next steps, open questions, all preserved as structured sections.
Rolling summaries
Older approved memory compresses into deterministic summary rows. Long-running projects stay readable without losing provenance.
API + MCP + SDK
Tools and agents read approved context through stable surfaces. Same trust contract on every surface.
Return with clarity.
Each project becomes a place you can resume from, with the context already loaded and a recommendation for which AI tool to open next.
Resume with context
Move work between ChatGPT, Claude, Claude Code, Cursor, Gemini, and Perplexity without rebuilding context. Each project generates a paste-ready context pack to copy or download as Markdown.
Next steps
Shelvia surfaces the moves worth taking next based on where the project actually stopped. Not generic advice.
Project timeline
A chronological view of imports, decisions, and saved prompts. Useful when reconstructing what happened.
Private by default
Each workspace is yours. Sharing is opt-in, scoped, and revocable.
The whole feature set is available on every plan that includes the category. Plan tiers shape how much you can capture and organize, not which capabilities exist.
See plan limitsCommon questions
If something here surprises you, email us at team@shelvia.net. We'll update the page.
Does Shelvia save AI output automatically?
No. Shelvia extracts useful suggestions from your imports and groups them for review. You decide what becomes project memory by approving, editing, or ignoring each suggestion. Your project memory should not be a black box; uncertain items are routed through review and the captured source stays visible where available.
Is this the same as saving AI chat history?
No. Chat history stores conversations in the order they happened. Shelvia is organized around projects. It connects prompts, notes, screenshots, links, files, decisions, and next steps so the useful work can be found and continued later. The raw chat stays archived; the structured pieces become memory.
What does "source-aware" actually mean?
When you save a passage, link, or decision, Shelvia keeps the original source attached: which import, which URL, which prompt produced what. Months later you can answer questions like "what was the citation behind that claim?" without trawling chat sidebars.
Do features depend on a specific AI tool?
No. Capture works with chats, screenshots, links, files, and notes from any tool. Organize and decide work on whatever you bring in. We don't require you to use any particular AI assistant.
Can I export everything?
Yes. Every project supports Markdown and HTML export. Your project memory is yours. Exportable when you want it elsewhere.
Does Shelvia train on my content?
No. Your imports, prompts, sources, and decisions stay inside your workspace. They are not used to train models. Workspace adaptation (better source-detection suggestions over time) is private to your workspace.
Your best AI work should not disappear into yesterday's chat.
Start with one chat, one screenshot, or one link. Watch it become memory you can return to.