Search, explore, and connect research papers through MCP. Built for LLMs. Open for everyone.
Papers With Code — the only structured database of AI/ML research — was shut down in July 2025. 500,000 papers and 1,500 benchmarks gone overnight.
Meanwhile, LLM agents need reliable access to scientific knowledge. Current APIs are rate-limited, lack full-text access, and none offer MCP integration.
No scientific publisher has built an MCP server. We did.
Connect your LLM to research papers in seconds. 4 tools: search, get_document, find_related, find_code. Hybrid search: BM25 + SPECTER2 + Gemini embeddings.
We don't just store PDFs. Each paper is decomposed into semantic chunks — individual ideas and claims — using LLM-powered analysis.
Open source. Open data. Open science. Apache 2.0 license. No paywalls. No vendor lock-in. A knowledge layer any AI application can build on.
Add to your MCP client configuration and start searching.
All MCP endpoints require a Bearer token in the Authorization header.
Tokens are issued through the OpenArx Portal.
After registration, generate an API key in your dashboard. The key format: oarx_sk_...
Each tool call deducts 1 credit from your balance. Check your balance via the Portal dashboard or via the get_my_profile tool.
| Tool | Description |
|---|---|
search | Hybrid search across papers — combines keyword and semantic matching |
get_document | Retrieve full paper details by arXiv ID |
find_related | Discover similar papers via vector similarity |
find_code | Find GitHub repos and datasets linked to a paper |