DocRouter AI: Open-Source Document Processing Data Layer
On This Page
DocRouter AI extracts, validates and routes structured data from operational documents such as invoices, manifests and insurance claims. It targets large, mixed document packets in regulated industries: healthcare, insurance, banking and supply chain. It runs as SaaS or self-hosted, and the source code is open under Apache 2.0.
How this profile came about: Andrei Radulescu-Banu requested the listing through the contribute form on 1 November 2026. IDP Software wrote the profile from DocRouter AI's website. Figures marked as vendor claims have not been verified independently. About the author: Andrei Radulescu-Banu.
Company and ownership
DocRouter AI is the trade name of Analytiq Hub LLC. The team page lists Andrei Radulescu-Banu as Founder and CEO and describes him as having "20+ years in LLMs, computer vision, fintech, digital health, self-driving cars, and Wall St infrastructure". The LinkedIn company page shows 2-10 employees and Boston, Massachusetts. The website gives no founding year, street address, funding or headcount, so none is stated here. Contact is through andrei@docrouter.ai, the "Contact us" link on the team page or the support page.
Where DocRouter AI fits
The product is a data layer between incoming documents and downstream systems such as ERP and workflow tools. The vendor says it can "split, classify, route, and reconcile 500-page medical, legal, or insurance packets". Under "Control cost at scale" the homepage says "Route easy work to inexpensive models". Industry pages cover banking, insurance, supply chain, clinical trials, DME and private equity.
Because the code is open, teams can run it themselves and change it; the vendor sells an Enterprise license with support, maintenance and upgrades. Compare it with other vendors in the IDP vendor directory. Insurance-focused alternatives include Indico Data and InEvo AI.
How DocRouter AI processes documents
The architecture page describes OCR through Amazon Textract (named as an example) and language models through LiteLLM, which can reach OpenAI, Anthropic, Bedrock, Vertex and Azure. Users define tags, prompts and schemas. The documentation also covers a Document Agent, Knowledge Bases (with a Knowledge Base Chat link) and Flows, a visual DAG editor and runtime. The insurance page lists "Advanced OCR for handwritten content with confidence-based flagging" and says "The system flags low-confidence fields for human review before transferring data to the AMS". This is classification and data extraction with configurable models.
Technical specifications
| Feature | Specification |
|---|---|
| Document Types | Invoices, manifests, insurance claims and applications (including ACORD forms), bills of lading (listed as a pilot), medical packets |
| Input Formats | Not listed in detail; multi-page packets up to about 500 pages |
| Output Formats | Structured data defined by schemas; webhooks and REST API |
| Processing Pipeline | OCR (Textract named), LLM extraction through LiteLLM, validation, routing, Flows |
| API/Integration | REST API, webhooks, Python and TypeScript SDKs, MCP server, N8N, Power Automate, Temporal |
| Deployment Options | Hosted SaaS; self-hosted with Docker Compose or Kubernetes |
| Certifications | None published |
| Claimed Accuracy | 99.9% precision (Boston Medical Data); 99% (banking target); no method published |
Deployment, security and pricing
The architecture page mentions encryption "in transit and at rest where configured for the deployment" and audit-oriented logging. It does not mention SOC 2, HIPAA or ISO 27001, so ask the vendor for any report. The SaaS pricing page lists Individual at $250 per month (5,000 SPU), Team at $1,000 per month (25,000 SPU) and Enterprise on request. New accounts start with 100 SPU credits. The vendor does not define SPU on that page.
Evidence
All results are DocRouter AI's own claims. The homepage states up to 90% faster processing, up to 95% error reduction and $50K+ annual savings without method or customer. The Boston Medical Data case study names the customer, for which DocRouter "processes high-volume peri-operative document packets for anesthesia revenue cycle management (RCM)" in "a live DocRouter deployment (on-prem)", and reports "175,000+ documents processed" and "99.9% precision on extractions in production". It gives no sample size, period or scoring method. The insurance wholesaler case study does not name the customer and calls itself "a completed pilot, not a production customer deployment"; it reports about 75% of clearance steps automatable and 90%+ accuracy on typed forms. The banking page labels 75% faster and 99% accuracy as target outcomes. No independent benchmark is published, so test the product on your own documents.
Resources
- Website
- Documentation
- Case studies
- Pricing
- Blog and talks
- Source code on GitHub (Apache-2.0)
- Atom feed
Company Information
DocRouter AI is operated by Analytiq Hub LLC, with the LinkedIn page listing Boston, Massachusetts. Andrei Radulescu-Banu is named as Founder and CEO. Address, founding year, funding and headcount beyond "2-10 employees" are not published on the website.