Quick answers.

Leapable gives your AI (Claude Code, Cursor, Cline, etc.) a private workspace it can search. You add source documents - PDFs, notes, transcripts, spreadsheets, images - and your AI can answer with cited passages from those sources. The workspace runs in the Leapable sidecar; PDFs, Office files, and images may use temporary cloud OCR/VLM processing unless an approved enterprise deployment keeps that processing inside your environment.

Create one workspace for the matter, then add productions, pleadings, transcripts, exhibits, and notes. Start with a focused ask like "show every admission about fall protection training". Leapable returns cited excerpts with document and page metadata so counsel can verify the support before using it in a memo, brief, or deposition outline.

Yes. Start with a representative folder, verify the citation quality, then batch the rest. Text and Markdown process locally. PDFs, Office files, and images may use cloud OCR credits, then extracted text and provenance return to your local workspace for search.

During processing, Leapable records the source file, page metadata, text offsets, source hashes, and provenance for each searchable passage. When your AI asks Leapable for evidence, the returned result includes the exact excerpt and page location, so you can click back to the source instead of trusting an unsupported answer.

Yes. Use a separate workspace for each matter, project, client, lab, or deal. Your AI searches the selected workspace by default. Cross-workspace search is available only when you explicitly ask for it, so unrelated material does not bleed into a narrow matter or project review.

Workspace mode is private: you use your own documents, keep full provenance, and verify answers against exact source pages. Creator publish is now Share Mode. It is optional and runs a sanitizer that seals a copy, removes sensitive source details, and lets subscribers ask against that published copy without receiving your raw files.

Private workspaces live on your machine by default, and text/Markdown files can stay fully local. PDFs, Office files, and images may use Leapable's cloud OCR unless an approved enterprise deployment handles that processing inside your environment. If you need a strict air-gapped or self-hosted path, contact support before adding regulated files.

No. You install it once (one-click .dmg / .exe / .deb), connect it to your AI (one command), then just talk to your AI. The How to use page has copy-paste prompts for every common task.

Realistic. Installer: ~3 minutes. Sidecar startup: seconds. Connect AI: 30 seconds. Create a workspace and add 10 markdown files: ~2 minutes. First ask: instant. PDFs take longer because of OCR.

Anything that speaks the Model Context Protocol (MCP). Confirmed working: Claude Code, Cursor, Cline, Continue, Claude Desktop, Warp. If a new client supports MCP, Leapable will work with it.

Not by default. Private text and Markdown files stay on your machine. PDFs, Office files, and images may go to Leapable's temporary OCR/VLM workers for text extraction, then the extracted text returns to your local workspace. If you explicitly publish through Share Mode, the published copy is sanitized and sealed before it is served.

They're uploaded to temporary signed storage in our datacenter, processed by our cloud OCR GPU worker, the extracted text is returned to your local Leapable sidecar, and the worker's storage is wiped. We don't store your PDFs or OCR output on our infrastructure.

Private workspace data is not used for model training. If you choose to publish through Share Mode, the legal flow asks you to reaffirm the copyright and AI-training terms for the sanitized, PII-stripped sealed copy before every publish.

Yes. Ask your AI: "delete the <name> Leapable workspace and all its documents, excerpts, semantic index, and provenance." It's a single tool call and it's permanent. You can also delete individual documents.

Ask your AI: "back up the <name> workspace to ~/backups/". Or use leapable_db_backup directly. The backups include all documents, excerpts, semantic index, and provenance.

PDF, DOCX/DOC, PPTX/PPT, XLSX/XLS, TXT, CSV, MD, VTT (WebVTT transcripts), and common images (PNG, JPG/JPEG, TIFF/TIF, BMP, GIF, WebP, HEIC/HEIF).

In practice, hundreds of thousands of documents. The local search store scales well, and retrieved answers stay tied to source excerpts so you can inspect the evidence instead of relying on a broad summary.

Leapable combines BM25 keyword search, a GTE multilingual semantic index, and a cross-encoder reranker. Beats single-method search by a wide margin. Every answer cites the exact excerpt + page it came from, so you can verify.

In your private workspace: yes — every retrieval carries the source document, page number, and exact excerpt. Citations are clickable in the dashboard and open the original at that page.

In Share Mode: answers are still grounded in the creator's actual source library and passed through the same reranker, but the Creator Publish Sanitizer strips filenames, paths, and raw-document links before publish. That's the creator's IP, protected by design — you get the answer, not the source files.

Yes — ask your AI for a cross-workspace search. It calls leapable_search_cross_db which semantic-matches across everything you've added and returns results ranked by relevance with the originating workspace.

New accounts include 5,000 starter credits for Library asks and cloud OCR, unlimited local processing for .md and .txt files, local semantic indexing, reranking, semantic search, and free/open Library content. No card required.

A credit is the usage unit for Library asks and cloud OCR. One credit covers one standard Library ask or up to 3 OCR page-equivalents, billed by run. Direct XLS/XLSX spreadsheet parsing is billed by data volume: 1 credit per started 1,000 rows or 100 KB, whichever is larger. Text-fast files (.md, .txt, .csv) process locally for free. 1 credit = 5¢ face value. Batch billing rounds up at the run level — a 6-page doc is 2 credits, a 24-page doc is 8 credits. A batch of 1000 files x 2 pages each is ceil(2000/3) = 667 credits, not 1000.

Subscribers get monthly credits included and 50% of unused credits roll over at cycle reset. One-off packs are exactly face-value (5¢/credit) with no rollover. If you ask regularly, a plan is meaningfully cheaper.

Yes. Upgrades take effect immediately — Stripe prorates the billing, and we grant prorated bonus credits for the rest of the cycle. Downgrades also take effect immediately, with unused time from the higher plan credited to your next invoice. Your current credit balance is untouched; the smaller monthly grant just applies on the next cycle reset.

Yes. Cancellation takes effect at the end of the current billing period — you keep access until then. No refunds on partial months. One-off credit packs never expire so long as you remain an active account.

All payments go through Stripe. We never see or store your card number. Stripe supports every major card, Apple Pay, Google Pay, and several regional methods. The "Secure Stripe" lockup on every checkout page is real — click it to verify.

Yes. Any workspace you've built can be published through Share Mode. You choose free or paid, per-ask pricing, and an optional monthly subscription. Leapable keeps 30%, you keep 70%.

Stripe Connect. You onboard once (takes ~5 minutes — ID, bank, tax). We pay out monthly on the 1st once your payable balance reaches $25. Workbench shows holding, reserve (5% / 60-day hold), and payout history.

Same as every payment processor — it covers chargebacks and refunds. After 60 days with no disputes, it releases to your holding balance automatically.

Yes. Each published workspace can have its own per-ask credit price (0-20 credits, N=0 means free/open-source). Optionally add one subscription per published workspace (e.g. $29/mo unlimited access).

No. Every publish runs through the Creator Publish Sanitizer (CPS) — a 12-phase pipeline that redacts PII via Presidio, anonymizes filenames to doc_<hash>, nulls source file paths and headings, drops identity tables, and seals the result with SHA-256 before upload. Subscribers receive answer text and anonymized excerpt IDs — not your raw files. There is no API that returns original documents.

Original filenames and paths, document titles, authors, headings, datalab IDs, per-excerpt provenance links, audit logs, annotations, saved searches, workflow state, user records, approval chains, and webhook configs. A final zero-residue regex sweep looks for emails, phone numbers, SSNs, credit cards, and person names — if anything slips through, the publish is blocked. You keep the canonical workspace on your machine; only the sealed, redacted copy ships.

Private workspace data is not used for model training. Published Share Mode copies are different: before every publish, you reaffirm the copyright and AI-training terms for the sanitized, PII-stripped sealed copy. If you do not want that grant to apply, keep the workspace private and do not publish it.

Yes. Unpublish at any time and the sealed copy is revoked from Share Mode. Uninstall the app and your local canonical data goes with it. No lock-in — the source of truth always lived on your machine.

Check the Leapable local sidecar is up: open your terminal and run curl http://localhost:4100/health. If that fails, restart the Leapable app. If it succeeds, restart your AI client and confirm the MCP is configured.

Common causes: (1) you are out of credits or the file is invalid — check your credit balance; (2) the PDF is password-protected — remove the password first; (3) the file is corrupted. Ask your AI: "retry failed docs in {name}".

Tell your AI: "rebuild semantic index for {name}". It'll re-embed the whole source library against the latest model version.

Platform subscription = the monthly fee, regardless of usage. Credits are consumed per ask/OCR page. If you want to stop being billed, cancel your subscription from Account & Plan. Unused credits stay on your account.

The How to use page has step-by-step prompts. For anything not covered, email support@leapable.ai — we answer within 24h.