Skip to content

Four lines to a cited passage.

Send any file in. Get back the passage, the file and the page it came from.

DocSlurp does three jobs: it extracts any file into page-by-page Markdown and structured fields, searches the result for RAG context, and answers questions through a configurable chat endpoint. Every passage and answer cites the page it came from. The CLI, SDK, HTTP API, MCP server and embeds all read and write the same projects.

Upload and ingestExtract and connectSearch and answerReferenceSelf-host

Three jobs, one project

Start with the job you need. Each one reads the same project, so documents you ingest are searchable and answerable without extra setup.

Ingest and extraction

Any file in: PDF, Office, EPUB, HTML, images, audio and video. Out come page-by-page Markdown, page images, chunks and, with your JSON schema, structured fields.

RAG search

Hybrid search returns ranked passages or ready-to-paste RAG context, each citing its document and page.

Configurable chat

Cited answers over a project, with streaming. Set instructions, answer style and context mode per project; pick the model, retrieval and filters per request.

Search your first file

You need Node.js 20+ and a DocSlurp account. Create a server key with the ingest and search:deep scopes, then run these four lines. Without a project ID, both commands use your organization's default project.

first-answer.shTerminal
npm install --global @docslurp/cli
export DOCSLURP_API_KEY='dslp_priv_…'

docslurp upload ./growth-plan-2027.pdf --watch
docslurp search 'Are we on track to beat our FY2027 earnings?'
  1. Line 1: Installs the docslurp command. The SDK and HTTP API do the same job from code.
  2. Line 2: A server key with the ingest and search:deep scopes. It stays on your machine or backend. To sign in with your browser instead, skip this line and run docslurp login --scopes "search:deep read ingest".
  3. Line 4: Without --watch, upload returns once the file is accepted. With it, the command follows processing until the run finishes, for up to five minutes.
  4. Line 5: Prints ranked passages, each with its file and page. Search doesn’t write an answer; docslurp chat does.

Outputdocslurp search 'Are we on track to beat our FY2027 earnings?'

1. [0.8732] growth-plan-2027.pdf (p.12)Revenue reaches $48.2M, 12% ahead of plan. 2. [0.7914] q3-earnings-memo.pdf (p.3)Q3 gross margin reached 71%, six points above plan.

FY2027 Growth PlanSection 4

4.3 FY2027 projection

Net revenue retention is 128%. Revenue reaches $48.2M, 12% ahead of plan. Margin lands at 24%.

12

growth-plan-2027.pdfp. 12 cited by result 1

Pass --context for numbered passages to paste into your own model, --json for the full response, or use docslurp chat for a generated answer with the same citations.

One project, six ways in and out

Upload from a script, search from your app, ask from your agent. Everything reads and writes the same projects, so you can mix them freely.

Your project

  1. Parse and OCR
  2. Chunk and embed
  3. Index with page numbers

Contents

Upload and ingest

  • Node SDK and CLI

    Upload, wait for indexing, and search from TypeScript or a terminal.

  • HTTP API

    Plain REST and JSON, from any language.

Extract and connect

  • Structured output

    Extract JSON fields with your own schema and check them against the source page.

  • Integrations

    Every SDK, adapter, CMS target and language client, with status.

Search and answer

  • Learn

    Illustrated explainers for every idea behind document search: OCR, embeddings, HyDE, ranking, citations.

  • MCP server

    Let Claude Code, Codex, Cursor or VS Code search your projects and cite pages.

  • OpenAI-compatible chat

    Point an existing chat client at a project.

  • Embeddable search

    Add search and chat with citations to your product with React components or custom elements.

  • Search and CMS guides

    Keep your Algolia, Pagefind, Meilisearch or Typesense code and swap the engine.

  • Algolia compatibility

    Keep your Algolia client calls and import settings, synonyms and rules.

Reference

  • SDK reference

    SDK methods, errors, retries and upload recipes.

  • Search reference

    Search inputs, response contracts and retrieval controls.

  • Chat tiers

    Pick fastest, balanced or smart; DocSlurp manages the models.

  • CLI reference

    CLI commands, JSON output and CI usage.

  • Packages

    Every package, with source and npm links.

Self-host

Jump to