Turn complex documents into context your agents can use.

Knowhere preserves structure, visual context, and source links, giving agents information they can navigate, retrieve, and verify.

Start free trialRead the docs

[ PRODUCT ]

See how Knowhere turns documents into structured, traceable context.

Explore page-level evidence, text content, and spreadsheet tables across PDF, Word, and Excel.View all supported formats

# pip install knowhere-python-sdk
from pathlib import Path
from knowhere import Knowhere

client = Knowhere(api_key="sk-YOUR_API_KEY")
documents = Path("nvidia-reports").glob("*.pdf")
results = [client.parse(file=file) for file in documents]
DOCUMENT 22026 NVIDIA Corporation Annual Review.pdf
SECTION 1AI Reinvents Computer Graphics

GeForce brought AI to the world. Now AI has come home to GeForce.

2026 NVIDIA Corporation Annual Review.pdf · PDF page 10 · AI Reinvents Computer Graphics
PAGE 10AI Reinvents Computer Graphics

Computer graphics is no longer limited by traditional rendering.

With neural rendering and DLSS, AI predicts pixels instead of computing them directly.

For every pixel we render, AI generates the rest in real time.

Neural networks now run alongside shaders and ray tracing cores. AI does not replace graphics. It enhances it.

In the era of neural rendering, graphics is learned.

Relationship
TABLE 01PDF · PAGE 01
GAAP$ in millions, except earnings per share
Q1 FY27Q4 FY26Q1 FY26Q/QY/Y
Revenue$81,615$68,127$44,06220 %85 %
Gross margin74.9 %75.0 %60.5 %(0.1) pts14.4 pts
Operating expenses$7,621$6,794$5,03012 %52 %
Operating income$53,536$44,299$21,63821 %147 %
Net income$58,321$42,960$18,77536 %211 %
Diluted earnings per share$2.39$1.76$0.7636 %214 %
TABLE 02PDF · PAGE 02
Revenue by Market Platform$ in millions
Q1 FY27Q4 FY26Q1 FY26Q/QY/Y
Data Center$75,246$62,314$39,11221 %92 %
Hyperscale37,86933,81417,59912 %115 %
AI Clouds, Industrial, & Enterprise37,37728,50021,51331 %74 %
Edge Computing6,3695,8134,95010 %29 %
Total$81,615$68,127$44,06220 %85 %
IMAGE 03PDF · PAGE 10
2026 NVIDIA Corporation Annual Review.pdf · PDF page 10 · AI Reinvents Computer Graphics
Source-backed context
{
  "project": "NVIDIA reports",
  "summary": "NVIDIA's Q1 FY2027 revenue reached $81.615B, up 85% year over year, with a GAAP gross margin of 74.9%. Data Center contributed $75.246B of revenue, led by hyperscale at $37.869B and AI clouds, industrial and enterprise at $37.377B; Edge Computing added $6.369B. The Annual Review's graphics example illustrates a separate application of AI: neural rendering and DLSS predict pixels alongside traditional rendering.",
  "citations": ["DOCUMENT 1 · PDF page 1", "DOCUMENT 1 · PDF page 2", "DOCUMENT 2 · PDF page 10"],
  "documents": 2,
  "source_regions": 3
}

[ PROCESS ]

A document pipeline that keeps context connected.

From ingestion to structured output, every step keeps the document connected to its original source.

01

Ingest documents

Upload PDFs, spreadsheets, presentations, scans, and other supported formats.

02

Capture every page

Extract text and visual regions while preserving the original page layout.

03

Map document structure

Identify headings, tables, formulas, layouts, and relationships across the document.

04

Return traceable context

Receive structured JSON, navigable document maps, and source-linked pages for agents.

[ CAPABILITIES ]

Built for documents that plain text cannot capture.

Preserve page structure, visual meaning, formulas, and source relationships across complex file formats.

Works with common document, presentation,
spreadsheet, image, and text formats:

DOCX, PDF, JPG, PPTX, XLSX, CSV, PNG, MD, JSON, and TXT.

Support for EPUB, HTML, XML, MP4,
MP3, and SKILL.md is coming soon.

Page-native understanding

Preserve layouts, diagrams, and spatial relationships so agents can inspect the original page when text is not enough.

Agent-ready structure

Turn flat text into hierarchical context that agents can scan first and expand when needed.

Formula recognition

Extract mathematical formulas and chemical structures in formats agents and downstream systems can use.

Source-level tracing

Trace extracted content back to its page and source region for easier verification.

On-premise deployment

Deploy Knowhere locally when document privacy, compliance requirements, or infrastructure control matters.

API-first integration

Connect Knowhere to existing workflows through REST APIs, webhooks, SDKs, or MCP.

[ COMPARISON ]

Compare document understanding across structure, accuracy, and traceability.

See how Knowhere and other document-processing tools compare across efficiency, accuracy, structure, and source tracing.

Performance comparison

Raw DocsUnstructuredKnowhereMinerUMarkitdown
Tokens usedProcessing timeAgent iterationsFirst-pass accuracyAccuracy after feedbackRecall
Capability matrix
Feature
Knowhere
Typical document parsers
Document hierarchy
Supported
Limited
Complex merged cells
Supported
Limited
Table boundaries
Supported
Not supported
Source tracing
Supported
Limited
Progressive context
Supported
Not supported
Visual understanding
Supported
Limited

[ INTEGRATION ]

Add document understanding to your workflow in minutes.

Submit a document through the API, receive structured results, and connect them to your existing agent workflow.

  1. 01Get an API key

    Sign up and generate your API key from the dashboard.

  2. 02Submit a document

    Send a URL or upload a file to the processing queue.

  3. 03Receive structured results

    Get structured JSON through webhook or polling.

# pip install knowhere-python-sdk import knowhere client = knowhere.Knowhere(api_key="sk_...") result = client.parse(url="https://arxiv.org/pdf/1706.03762.pdf") print(result.statistics.total_chunks) print(result.full_markdown[:200])
Illustrative flow · interface details unconfirmed
# Illustrative only — no real endpoint
result = knowhere.process("sample.pdf")
print(result.structure)
// Illustrative only — no real endpoint
const result = await knowhere.process("sample.pdf")
console.log(result.structure)
# Illustrative only — no real endpoint
curl -X POST "[endpoint-to-be-confirmed]" \
  -F "file=@sample.pdf"

Use Knowhere through MCP

Bring structured document context into Cursor, VS Code, Claude, or Codex.

Read the MCP docs

[ PRICING ]

Simple, transparent pricing.

Pay only for what you use. No hidden fees, no complex tiers.

Estimated cost ($0.015 per page)$9.00
Estimated budget
$9.00
100-page PDFs
6 documents
500-page documents
1 document
Commitment
No minimum
$9.00
100 pages2,5005,0007,50010,000

File Size Limits

Need higher limits? Contact team@knowhereto.ai
for enterprise pricing with custom limits.

.pdf
100M
.docx
50M
.xlsx
100M
.pptx
100M

[ ENTERPRISE ]

Need custom limits or deployment support?

Custom limits, deployment, support, and SLAs.

Talk to our team
  • Custom rate limits

    Discuss throughput limits for your production traffic.

  • Priority processing

    Plan priority handling for time-sensitive workloads.

  • Deployment options

    Review managed, dedicated, or self-hosted options.

  • Support requirements

    Align support channels and response expectations.

  • SLA requirements

    Define your uptime and service-level needs.

  • Commercial terms

    Discuss billing and terms for your usage.

[ FAQ ]

Frequently asked questions.

When am I charged?

Page credits are deducted when a job completes successfully. Failed jobs do not consume credits.

Page credits expire 3 months after purchase.

Contact team@knowhereto.ai for refund requests within 14 days of purchase.

We accept all major credit cards through Stripe: Visa, Mastercard, American Express, and more.

[ GET STARTED ]

Ready to build with better document context?

Connect your agent workflow and see how Knowhere handles documents that plain text pipelines miss.

  • Free 14-day trial
  • No credit card required
  • Cancel anytime