Drop in a PDF, a contract, a whole handbook. Inkwell reads it cover-to-cover so you don't have to. Get answers in seconds — with page-level citations.

PDFs, Word docs, plain text. Drag and drop or click upload. We chunk and index automatically.
Server-side parsing, smart chunking, vector embeddings. Indexed in seconds, not minutes.
Natural language questions. Cited answers with page references. Conversational follow-ups.
Every answer links back to the exact passage. No more “trust me bro” AI.
Average response under 3 seconds. We don't make you wait around.
Workspace-scoped storage. Row-level security. Delete anytime, no questions.
Ask questions that span dozens of files. Inkwell connects the dots.
Follow-up questions remember context. Like talking to a research assistant.
Semantic splitting preserves meaning across page breaks and tables.
“Replaced an entire afternoon of Ctrl+F. I just ask. Wild.”
“Our onboarding doc is 240 pages. New hires now actually use it.”
“Citations are clutch. I can verify everything before I act.”
“We dumped 6 years of research papers in. It just… works.”
“Faster than asking my coworker. Less awkward too.”
“Killed three internal tools. Inkwell does the job of all of them.”
PDF, DOCX, TXT, Markdown, and most plain-text formats. Scanned PDFs are OCRed automatically.
Up to 50MB per file on the free plan. Bigger files? Reach out — we can help.
In your own workspace inside Supabase, with row-level security. Only people you invite can see it.
Every answer ships with page numbers and the exact passage we pulled from. No hand-wavy AI.
Yes — one click. We purge the file, the chunks, and the embeddings.
A mix. Gemini for fast retrieval, Claude for nuanced reasoning, OpenAI for embeddings.
Workspaces support inviting members with role-based access. Owners, admins, members, viewers.
ChatGPT guesses. Inkwell reads your actual documents and cites them. Big difference.