Turn your documents intogrounded answers.

Graphor is a document intelligence platform. Upload files, web pages, repositories and media; an agent investigates them the way a person would — narrowing files, searching semantically and by regex, reading pages, checking charts — and answers with citations. Dashboard, REST API, SDKs and MCP.

Trusted by teams at

One platform from raw documents to grounded answers.

Graphor covers the full document intelligence pipeline as a managed platform: you bring the documents and consume the results through the dashboard, the REST API or the SDKs — no retrieval infrastructure to build or operate.

Multi-source ingestion

25+ file types plus web pages, GitHub repos and video — five parsing methods, from fast heuristics to agentic OCR, with Auto routing every page to the cheapest one that can read it.

Agentic search

Not one-shot retrieval: the agent narrows the files, runs semantic and regex search, opens the pages that matter and sends charts to a vision model — for as long as the question needs.

Grounded answers

Every claim carries the file, page and passage behind it; contradictions are resolved by following references, not averaged away. Effort levels budget how deep each question searches.

Structured extraction

Define a JSON Schema and extract typed data across whole batches with page-level provenance — and export the result as CSV or markdown.

Developer API & SDKs

REST API, TypeScript and Python SDKs, and a 12-tool MCP server so your coding agent queries your knowledge base directly. Everything in the dashboard is an endpoint.

Versioned builds & cost control

Every ingestion is a versioned build: compare parsing methods, switch the active version instantly, and skip enrichment or indexing when a run only needs parsing.

What happens inside the platform.

Every source follows the same observable pipeline: each step produces an artifact, every output carries its source, and every build is versioned.

sources.ingestFile()step 01 / 6

Ingest

Upload a file or point Graphor at a URL, repository or video. Ingestion runs asynchronously with per-page progress you can poll.

Artifact: a versioned build with status you can track.

The platform, and the vertical built on it.

Graphor DocumentAI is the core platform. Legal.Ops is the first vertical product built on top of it — proof of what the same APIs let you build.

Building something else on top of documents? The API is the product — talk to us.

From first upload to production.

No procurement cycle, no implementation project. Create a project, ingest your documents and query them the same day — then ship it to your users through the API.

01Create a project

Each project is an isolated knowledge base with its own indexes, API keys and usage metering.

02Ingest sources

Upload files or connect URLs, repositories and videos — via dashboard or API. Pick the parsing method per source.

03Query & evaluate

Ask, extract and retrieve over your real documents. Inspect every parsed element and every citation before you ship.

04Ship & scale

Integrate with the SDKs or MCP server. Usage-based billing scales with what you process, not with seats.

Private by project

Your documents, indexes and extractions are isolated per project and queryable only with your keys. Your data is not used to train models.

Verifiable answers

Every answer and every extracted field links back to the source passage it came from. Trust is checkable, not claimed.

Versioned builds

Every ingestion is a versioned build with full history: compare versions, roll back, retry failed pages, or re-index without re-parsing.

What teams run on Graphor.

Sales operations

BD & SDR orchestration agents

Lead classification and prioritization, conversation handoff and routing, CRM integration and AI-driven follow-ups — retrieval grounded in Graphor.

3.2×
pipeline coverage
Customer experience

Support copilots on internal knowledge

Copilots grounded in product docs and past tickets resolve routine cases and draft the rest for human review, citing their sources.

−58%
first-response time
Health & marketing

Conversation intelligence & analytics

Chat, email and Slack conversations turned into sentiment, trends and engagement insights in real time, through the extraction API.

6
channels analyzed in real time
Legal

Case intake automation (Legal.Ops)

Scattered email threads and attachments become consolidated cases with state, owner, deadline and verifiable sources — the vertical built on the platform.

11 hrs
saved per case, avg.

Your software is blind to 80% of your data.

Most company knowledge lives in documents no dashboard reads: contracts, threads, transcripts and PDFs. Retrieving it means asking whoever was there — or building a document pipeline from scratch.

Engineering teams
Months
building document pipelines by hand

OCR, chunking, embeddings, vector stores, rerankers, evals: a full retrieval stack before the first useful answer ships.

The data
80%
of company data is unstructured

Contracts, threads, transcripts and PDFs that no dashboard reads. The knowledge exists; software just cannot query it.

AI products
Zero trust
without verifiable sources

LLM answers nobody can check do not survive contact with legal, finance or operations. Grounding with citations is the difference between a demo and a product.

Common questions.

A document intelligence platform. You ingest unstructured sources — files, web pages, repositories, media — and an AI agent investigates them: it searches semantically and by regex, reads the pages that matter, checks charts with a vision model and answers with citations. Available via dashboard, REST API, TypeScript/Python SDKs and MCP.

No. One embedding query returns whatever is nearest and hopes it is right. Graphor runs a real tool loop: it narrows the files, combines semantic search with exact regex matching, reads tables by row and header, sends charts and scans to a vision model, and resolves contradictions by following references between documents. Provenance is page-level on every claim.

Access is currently rolling out in waves: request access and we open your account with credits to evaluate on your own documents. Existing users log in at app.graphorlm.com.

Usage-based credits: you pay for the processing you run — parsing by method and volume, queries and extractions — not per seat. Per-ingestion controls let you skip enrichment or indexing for cheaper, faster runs.

Each project is an isolated knowledge base: your documents, indexes and extractions are private to it and reachable only with your keys. Your data is not used to train models. Details are in the trust section of our docs.

PDFs, Office documents, spreadsheets, images, audio and video, plus web pages, GitHub repositories and YouTube videos. Parsing methods range from fast text heuristics to OCR and agentic parsing for complex layouts — including an auto mode that routes each page.

REST API with access keys per project, official TypeScript and Python SDKs, and an MCP server that plugs the platform into agent frameworks and AI-native tools. Everything in the dashboard is available through the API.

Graphor DocumentAI is the horizontal platform: ingestion, retrieval, Q&A and extraction as APIs. Legal.Ops is a vertical product built on top of it for legal back-office automation. Same engine, different packaging.

Who builds Graphor.

Graphor is built by a Brazilian product engineering team obsessed with one problem: making unstructured data usable by software. The platform started as the retrieval engine behind production AI systems and became a self-service product — the same APIs now power Legal.Ops and our customers’ own applications.

Lucas Neves

Co-founder

LinkedIn

Thiago Hirano

Co-founder

LinkedIn

Rodrigo Lima

Co-founder

LinkedIn

Ready to query your documents?

Request access and evaluate Graphor on your own documents — or write to us about volumes, API licensing and enterprise plans.