Outerport Logo

Make every diagram, drawing, and document queryable.

Turn every diagram, drawing, webpages, documents, CAD files and more into an exact text representation that can be ingested by LLMs, or into an API to be used by AI agents. Ship enterprise search and agents that don't just end with text data.

Trusted by Fortune 500 companies and financial institutions

Visual-Diagram Parsing

We use a combination of computer vision and VLMs to parse visual diagrams into code, which can reproduce the original diagram and expose a set of APIs for agents to use.

Text (code)-based question and answering is significantly more accurate than image-based question and answering, while being 30x cheaper.

Benchmark comparison showing better benchmarks fow flowchars

Original Flowchart (Image)

Original flowchart

Parsed Output (Code)

Parsing results

Enterprise-grade Search

Our agentic search system improves upon vector RAG and naive long context approaches, delivering more accurate and faster retrieval through advanced techniques like KV cache management and chain-of-memory reasoning.

Read a detailed blog post about our tech
Benchmark comparison showing superior accuracy

Agentic Retrieval

Outerport uses an LLM-native approach to retrieval instead of using vector embeddings like most RAG systems. Unlike vector RAG, Outerport handles complex queries that need reasoning or comprehensive answers.

Example: "Flag all sections that violate compliance guidelines" can't be expressed as vector similarity.

10-20x More Efficient

Our "KV-cache management" technology makes processing very long documents dramatically more resource efficient than traditional approaches, without sacrificing quality.

Get the same powerful reasoning capabilities at a fraction of the usual cost through optimized memory management.

Easy API integration for customization.

from outerport import OuterportClient

# Initialize with your API key
client = OuterportClient(api_key="...")

# Upload a document
with open("policy.pdf", "rb") as f:
  client.documents.create(file=f)

# Find a document
docs = client.documents.search(query="Policies on information security")

# Ask a question about the document
question = client.questions.create(
    documents=docs,
    question="What restrictions exist on the use of portable storage devices?",
)

# See the evidences
print(question.evidences)
Python
TypeScript

Get access
immediately.

Trusted by Fortune 500 companies and financial institutions. Built by AI engineers and research scientists from NVIDIA, Meta, LinkedIn,

Contact us at: info@outerport.com

Frequently Asked Questions