Octogen
Menu
For BrandsFor AgenciesFor DevelopersDocs
Sign in

For Developers

Commerce databuilt for agents.

Search, lookup, brand discovery, and more-like-this on top of a deeply enriched product graph.

Read the docsExplore Cosimo

A single API for search, lookup, brands, and MLT.

API explorerClient surface
REST

One client surface

Search, lookup, brands, MLT

Semantic search

/v1/search

GET

Product lookup

/v1/products/{product_id}

GET

Brand discovery

/v1/brands/{brand_slug}

GET

More like this

/v1/products/{product_id}/similar

GET
SDK consoleoctogen.ts
TypeScript
octogen.ts
const results = await octogen.search({
  query: "waterproof trail jacket",
  filters: {
    material: "recycled nylon",
    activity: "hiking"
  },
  include: ["brands", "more_like_this"]
});

products

brands

similar

Standardized deep schema and enriched ontology covering billions of products.

Octogen normalizes messy storefront data into consistent product, variant, image, price, attribute, taxonomy, and ontology structures designed for agentic commerce.

Schema explorerProduct ontology
Normalized graph

Products

Variants

Images

Attributes

Enriched ontology

taxonomy pathmaterial familiesfitoccasionpatternaudiencestylecompatibility

Agents specialized in ecommerce data discovery, extraction, restructuring and enrichment.

Ecommerce data is notoriously messy.
Since we can't --yet-- afford "To organize the world's information and make it universally accessible and useful", we trained our agentson thousands of ecommerce sites instead.
Use Voyager to discover, extract, structure and enrich any catalog into a consistent standardized and rich schema.

Voyager workflowCatalog discovery, extraction, structure and enrichment
Agent run
01

Discovery

Deep site analysis

Is it ecommerce?
Find product URLs
Sample product URLs
Platform-level quirks
Data Source Discovery
Consolidate data sources
02

Data collection

Collect product data and analyze quality

Process Sitemaps
Process product page samples
Analyze data quality
Prepare LLM input
03

Static extraction

Run LLM extraction, reconcile discrepancies, build ground truth. O(1) solution.

Run LLM extraction
Reconcile discrepancies
Build grounded evidence
Reconstitute product groups
Compute consensus
Build *Ground Truth*
Optimize image URLs
04

Code Generation

Generate and validate extractor code. O(N) solution

Evaluate existing extractors
Run coding agent
Judge against ground truth
Iterate with corner-cases
Optimize performance
Run Extraction
05

Promotion

Evaluate extraction quality and production promotion

Auto-merge pull request
Promote to worker pool
Detect data drift
Apply enrichments

voyager.discover("new-storefront.com")

Blazingly fast responses and CDN cached images.

Low-latency APIs and stable CDN image URLs keep agent workflows, product discovery, and visual experiences responsive at production scale.

Performance monitorRuntime paths
Live

Built for production paths

Search, PDPs, agents, and visual UIs

API responses
Fast reads
Image delivery
CDN cached
Agent payloads
Compact JSON
CDN contractImage delivery
Stable URLs

Cached image contract

cdn.octogen.ai/images/

catalog/product/primary.webp

webpresizestable