Start by role: vertical quick-starts

Pick the closest role, then ask your AI client to run the MCP command sequence. Each path ends with a cited source check so the answer can be verified against the original files.

Litigation

I'm a litigator — drop production folder, ask cited questions

Use the litigation workflow when a production, transcript set, or brief bank needs page-cited answers. See the litigation page.

Create one legal-model vault per matter.
Add the production folder recursively and let processing finish.
Ask a full natural-language question that names the issue and desired citation.
Open the cited source page before relying on the answer.

leapable_db_create name="acme-production" tags=["litigation","matter"] embedding_model="legal"
leapable_ingest_directory directory_path="~/Matters/Acme/Production" recursive=true
leapable_search query="Which produced documents discuss fall protection training admissions and who made those statements?" mode="hybrid" include_provenance=true
leapable_document_page document_id="{document_id}" page_number=649

Screenshot placeholder: production folder added, cited result, source page.

Research

I'm a researcher — add Zotero export, sentence-cited synthesis

Use this for paper libraries, lab notes, and literature reviews that need sentence-level citation trails. See the research page.

Create a vault for the review topic.
Add the Zotero export plus the linked PDF folder.
Ask for a synthesis that names the claim type and citation requirements.
Export or inspect the cited passages before using them in the review.

leapable_db_create name="zotero-sleep-review" tags=["research","zotero"]
leapable_ingest_files file_paths=["~/Zotero/Exports/sleep-review.csv"]
leapable_ingest_directory directory_path="~/Zotero/storage" recursive=true file_types=["pdf"]
leapable_search query="What does the literature say about sleep restriction effects on working memory, with sentence-level citations?" mode="hybrid" include_provenance=true

Screenshot placeholder: Zotero sources, synthesis answer, cited sentence trail.

Deal team

I'm on a deal team — connect VDR, generate IC memo

Use this when a VDR export, CIM, model, or diligence checklist needs a memo with source-backed figures. See the deal team page.

Sync or export the VDR to a local folder the Leapable runtime can read.
Create a deal vault and add the full VDR directory.
Ask for red flags, financial claims, and diligence gaps with source links.
Build the IC memo only from cited passages and tables.

leapable_db_create name="northstar-vdr" tags=["deal","diligence"]
leapable_ingest_directory directory_path="~/Deals/Northstar/VDR" recursive=true
leapable_search query="Which VDR documents support the revenue growth claim and what diligence risks should appear in the IC memo?" mode="hybrid" include_provenance=true
leapable_document_tables document_id="{document_id}"

Screenshot placeholder: VDR import, risk table, memo-ready cited figure.

Claims

I'm a claims adjuster — add claim file, treatment timeline

Use this for policy PDFs, medical bills, photos, notes, and correspondence that need a defensible chronology. See the claims page.

Create one vault per claim number.
Add the claim folder, including scans, photos, and correspondence.
Ask for a dated treatment timeline with the source for each event.
Open the cited documents before sending a coverage or reserve note.

leapable_db_create name="claim-24-1187" tags=["claims","timeline"]
leapable_ingest_directory directory_path="~/Claims/24-1187" recursive=true
leapable_search query="Create a treatment timeline for this claim with dates, providers, injuries, and citations for every event." mode="hybrid" include_provenance=true
leapable_provenance_verify document_id="{document_id}"

Screenshot placeholder: claim folder, dated timeline, provenance check.

Journalism

I'm a journalist — process FOIA production, find buried entities

Use this for FOIA productions, emails, attachments, and public-record dumps where the first pass is entity discovery. See the investigators page.

Create a vault for the request or investigation.
Add the FOIA production folder and keep the directory structure intact.
Ask for people, organizations, dates, and unexplained references.
Verify each lead against the original page before reporting it.

leapable_db_create name="foia-water-board" tags=["foia","investigation"]
leapable_ingest_directory directory_path="~/FOIA/WaterBoard/Production" recursive=true
leapable_search query="Which people, organizations, dates, and unexplained references appear repeatedly across this FOIA production?" mode="hybrid" include_provenance=true
leapable_document_list limit=20

Screenshot placeholder: FOIA batch, entity leads, cited original email.

Founder

I'm a founder — add Drive, ask company docs

Use this for shared-drive exports, policies, product docs, sales notes, and meeting records that need one company answer surface. See the teams page.

Sync Drive or export the folders your AI is allowed to read.
Create a company vault and add the approved folder tree.
Ask operational questions with enough context for a cited answer.
Use cross-vault search when departments stay in separate vaults.

leapable_db_create name="company-drive" tags=["founder","team"]
leapable_ingest_directory directory_path="~/Google Drive/Company" recursive=true
leapable_search query="What do our onboarding documents say about enterprise customer handoff responsibilities after contract signature?" mode="hybrid" include_provenance=true
leapable_search_cross_db query="Which team vaults mention enterprise customer handoff responsibilities after contract signature?" max_total_results=8

Screenshot placeholder: Drive export, company answer, source handoff note.

Developer

I'm a developer — connect MCP, build agentic-RAG

Use this when Claude Code, Cursor, or another MCP client should ground agent work in repo docs, API notes, and runbooks. See the developer page.

Confirm your AI client can see the Leapable MCP server.
Create a docs vault for the repo or service.
Add markdown, OpenAPI specs, design docs, and runbooks.
Ask the agent to cite the source before changing code.

leapable_status
leapable_db_create name="api-docs" tags=["developer","rag"]
leapable_ingest_directory directory_path="~/repos/acme-api/docs" recursive=true file_types=["md","txt","json"]
leapable_search query="Which API runbooks describe how to rotate production webhook secrets without downtime?" mode="hybrid" include_provenance=true

Screenshot placeholder: MCP connected, docs vault, cited runbook answer.

1. Install Leapable (one click)

Download the installer for your OS. Double-click it. That's the whole thing.

Download installer

What the installer does for you — automatically

You don't need to edit config files, run terminal commands, or configure MCP. The installer handles the local setup flow:

Verifies the bundled native runtime. The installer checks host support, disk space, and sidecar startup without requiring Docker, Node, or Python.
Installs the Leapable sidecar. The bundled Bun runtime handles local semantic indexing/search helpers. OCR and VLM models stay in Leapable's cloud OCR worker image.
Starts the MCP server on localhost:4100. Stores returned search vectors locally for fast private search. Neural conversion and result ordering run on Leapable's secure processing backend, not on your machine.
Auto-configures 10 MCP client configurations. If Claude Desktop, Claude Code, Codex (ChatGPT), Cursor, VS Code, Windsurf, Continue.dev, Zed, JetBrains, or Amazon Q is installed on your machine, the installer registers Leapable as an MCP server in their configs. You don't touch a JSON file.
Creates a persistent local data directory. Your vaults survive app and sidecar updates.
Auto-updates in the background. New signed app and sidecar bundles install without touching your vault data.

After the installer finishes, restart any running AI client (Claude Code, Cursor, etc.) so it picks up the new MCP server. That's the only manual step.

2. Verify your AI can see it

Open your AI and ask:

Use leapable_status to check that the Leapable MCP is healthy, and list any vaults I already have.

If your AI replies with status: ok and 191 tools available, you're done. If it can't find any Leapable tools, jump to Troubleshooting.

3. Create your first vault

Think of a vault as a folder your AI can read instantly. One vault per topic is a good default — "my-research-papers", "company-wiki", "contracts-2026", etc.

1Ask your AI to create a vault:

Create a new leapable vault called "research-papers". Give it the description "Academic papers I'm reading for my thesis on quantum networks" and tag it with "academic" and "physics".

2Your AI will call leapable_db_create behind the scenes. Confirm it worked:

Show me all my local leapable vaults.

Your AI will run leapable_db_list and show you a table with document counts, sizes, and last-accessed timestamps.

Pro tip: The description is what other creators' vaults get matched against when someone searches. Write it like a one-sentence pitch, not a folder note. Good: "Alex Hormozi's sales + offer frameworks from his book and podcast." Bad: "my docs".

4. Add sources

Leapable supports PDFs, DOCX, Markdown, plain text, images (via OCR), audio transcripts, and more. Just tell your AI where the files are.

From a folder on your disk

Select the leapable vault "research-papers", then add every .pdf and .md file from ~/Documents/thesis/papers — recursively into subfolders. When done, process anything pending.

Specific files

Add these three files into the "research-papers" vault: ~/Downloads/entanglement-2024.pdf, ~/Downloads/decoherence-review.pdf, ~/notes/chapter-3-draft.md. Process them after.

Watch the progress

Show me the processing status of the research-papers vault. How many docs are pending vs. complete?

Text-fast files (.md, .txt, .csv) process instantly on your machine. PDFs and images use cloud OCR (1 credit per 3 OCR page-equivalents; direct XLS/XLSX spreadsheets bill by started 1,000 rows or 100 KB). Everything else — source parsing, semantic indexing, search indexing — happens locally for free.

New accounts start with 5,000 credits, so you can process a meaningful first batch before buying more.

5. Search & ask questions

This is where it gets magical. Your AI can semantic-search (meaning-based), keyword-search (exact-word-based), or hybrid-search (best of both) across everything you've added.

Ask a question, get cited answers

In my "research-papers" vault, what do different authors say about decoherence in superconducting qubits? Give me the best 5 passages with citations.

Exact-phrase lookup

Search research-papers for the exact phrase "T1 relaxation time". Show the 10 most relevant hits and tell me which documents they came from.

Compare two documents

Compare the two papers on entanglement swapping in my research-papers vault. What do they agree on? Where do they disagree? Cite the specific sections.

Rolled-up summary of the whole source library

Summarize what's in my research-papers vault overall. Themes, how many docs, what areas are well-covered vs. thin. Keep it under 300 words.

Every answer carries citations. When your AI answers, it quotes the source document + page + exact excerpt. Click any citation in the dashboard to open the original document at that page.

6. Navigate & organize

Everything has a web view. Go to your Workbench to browse visually — or keep asking your AI.

List what's in a vault

Show me the first 20 documents in research-papers, sorted by most recent. Include file name, page count, and status.

Group similar documents

Cluster the documents in research-papers by topic. How many clusters did it find? Show me the top themes in each.

Add tags

Tag the research-papers vault with "thesis-2026" and "chapter-3-source".

Export everything

Export research-papers as a single JSON file I can back up.

7. Search across vaults

Have five vaults? Ask the question to all of them at once. Your AI routes the ask to whichever ones are likely relevant.

Across all my leapable vaults, find everything I have on "risk management in early-stage startups". Rank by relevance, show which vault each hit came from.

Under the hood your AI calls leapable_search_cross_db — semantic similarity across everything you've added.

Browse creator vaults in Discover

Discover lists vaults other creators have published — medicine, legal, finance, specific author catalogs, etc. You can ask them too (1 credit per ask).

Search Leapable Discover for vaults about "sales frameworks". List the top 5 — who made them, what's inside, and how much per ask.

8. Share or sell your knowledge

Any vault you've built can be published to Discover. Keep it free, set a per-ask credit price, or add a monthly subscription. Leapable keeps 30%, creators keep 70%.

Publish as-is

Publish my research-papers vault to Leapable Discover. Free for anyone to ask. Don't add a monthly subscription.

Publish with a subscription

Publish my trading-strategies vault with a $29/month subscription. 2 credits per ask. Give it a catchy creator headline for the Discover listing.

Manage your creator profile + Stripe payouts from Workbench.

9. Ready-to-use AI prompts

Copy-paste these into Claude Code / Cursor / Cline. Replace {placeholders} with your own values.

Researcher workflow

Create a leapable vault called "{topic}-research" tagged with "research". Add every PDF in ~/Downloads/papers/. Once processed, tell me how many docs, how many pages, average quality score, and any files that failed OCR.

Literature review

For my {topic}-research vault: (1) cluster all docs by sub-topic, (2) pick the 3 clusters with the most docs, (3) write a 200-word literature-review paragraph for each cluster, citing specific passages.

Contract review

Add ~/Documents/contracts/vendor-2026-Q2.pdf into a vault called "vendor-contracts". After processing, find every clause about indemnity, liability caps, and termination rights. Quote exact language and tell me the page.

Meeting notes knowledge-vault

Add every file in ~/Dropbox/meetings/ into a vault called "team-meetings". Once done, I'll ask you for decisions, action items, and follow-ups by person.

Code documentation index

Create a leapable vault "api-docs" and add every .md file under ~/repos/{project}/docs/. I want to ask it questions about our internal APIs.

Due-diligence dossier

Add everything in ~/diligence/{company}/ — PDFs, spreadsheets, PPTs — into a vault called "{company}-diligence". Then: (1) summarize their financial position, (2) list every risk factor mentioned anywhere, (3) compare their stated valuation against the assumptions in their financial model.

Personal archive / second brain

Walk ~/Dropbox/notes/ recursively and add everything into a vault called "second-brain". After that I want to ask it what I was thinking about on specific dates, or pull quotes from my old journal entries by topic.

10. MCP client setup details

Leapable speaks the Model Context Protocol (MCP), so anything that supports MCP can talk to it.

Claude Code

claude mcp add leapable -- ~/.leapable/bin/leapable-mcp --transport stdio — then restart. (AI clients connect over stdio; the local :4100 HTTP endpoint is the dashboard's browser-only back-channel.)

Cursor

Settings → MCP → Add Server → paste the JSON from section 1.

Cline / Continue

Both support MCP over stdio. Point at the native sidecar ~/.leapable/bin/leapable-mcp --transport stdio in their MCP config.

Claude Desktop

Settings → Developer → MCP Servers. Stdio transport via the installer-managed native sidecar ~/.leapable/bin/leapable-mcp --transport stdio.

11. Troubleshooting

My AI says it can't see leapable tools

1. Make sure the Leapable local sidecar is running: curl http://localhost:4100/health — should return {"status":"ok","tools_count":191}.

2. Restart your AI client after adding the MCP server.

3. Ask your AI: "list the tool namespaces you have available" — you should see a leapable_* namespace.

Source Processing looks stuck

Ask: "leapable retry failed documents in the {name} vault, then process pending". That resets stuck items and restarts OCR.

Search returns nothing

Ask: "use leapable_db_stats on {name} to tell me how many excerpts and semantic index exist". If excerpts=0 the documents haven't been processed yet. If semantic index=0 the GPU/CPU embedder couldn't reach the model.

I need help that's not here

Email support@leapable.ai or check the FAQ.