Maintenance mode — stable, in-production primitive
Any webpage.
Clean markdown. Under 500ms.
NeoReader converts web content into LLM-ready markdown. Built for ingestion pipelines — fast, predictable, no rendering artefacts. An ingestion primitive, not a scraping toolkit.
At a glance
Web-to-Markdown in Milliseconds
- Status
- Functional — maintenance mode (no active feature dev)
- Built for
- AI engineers and data teams building RAG pipelines or LLM ingestion flows
Why it matters
Positioning pillars
Speed at ingestion scale
Sub-500ms p95, at scale.
Designed for RAG pipelines that ingest thousands of URLs per batch. No full-browser tax on clean HTML; headless fallback only when the page demands it. Scales to millions of requests with 99.9% uptime as the operational target.
Framework-aware rendering
Handles React, Next, Vue.
Most sites return useful HTML on the first request — NeoReader converts them in milliseconds. JS-heavy sites (React, Next.js, Vue, Svelte SPAs) fall back to a headless rendering pipeline so the final markdown reflects what a user actually sees, not an empty shell.
Transparent, predictable output
Structured output, every time.
Clean markdown with title, body, and metadata in one response. Boilerplate and ads removed. No surprise HTML entities, no stale cached fragments, no tracking pixels. The output a RAG pipeline can index without a post-processing step.
The mechanics
How it works
Step 1
POST a URL
Send a URL to the NeoReader API (or call the MCP server). Receive structured markdown, title, and metadata in a single response — no multi-step crawling, no manual parsing.
Step 2
Fallback only when needed
Static or lightly-rendered pages return in milliseconds. JS-heavy pages automatically fall back to a headless pipeline — slower, but comprehensive. You pay for rendering only when the site requires it.
Step 3
Drop into your RAG pipeline
Pipe the markdown straight into GnosisLLM, a vector DB, or a chunker. NeoReader is an ingestion primitive — it ends where your knowledge pipeline begins.
Proof in production
<500ms p95
Sub-500ms p95 latency, 99.9% uptime target
Operational since 2025 as the web-ingestion backbone for GnosisLLM collections. Hundreds of thousands of URLs processed. Every Neomanex knowledge ingestion that starts from a webpage goes through NeoReader first.
NeoReader production metrics
The ecosystem
Fits into the portfolio
Related products
Questions, answered