OpenArx
@OpenArx-AI
About OpenArx
Open scientific knowledge MCP for AI agents. Three profiles: search (15 tools incl. find_evidence, compare_papers, explore_topic), publish (5 tools for direct submission with AI-assisted
Config
Add this server to your MCP-compatible client using the configuration below.
{
"mcpServers": {
"openarx": {
"type": "http",
"url": "https://mcp.openarx.ai/researcher/mcp",
"headers": {
"Authorization": "Bearer oarx_sk_YOUR_TOKEN_HERE"
}
}
}
}Tools
No tools detected
We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.
Overview
What is OpenArx?
OpenArx is open infrastructure that lets an AI agent work from real research instead of guessing. Instead of answering from what a model happens to remember, your agent searches a live corpus of scientific papers, checks claims against it, and shows you what actually supports a statement and what contradicts it.
Papers are broken into idea-level units at ingest, so an agent works with individual ideas rather than raw PDFs. On top of that sits a graph of claims and the typed relations between them: what supports, extends, qualifies, or refutes what.
How to use OpenArx
Register at https://portal.openarx.ai to get an API key (format oarx_sk_...). Paste the server config into your MCP client, replace the token placeholder with your key, and connect. The role you choose when creating the token decides which tools your agent gets.
Key features
- Semantic, keyword, and hybrid search across a corpus of scientific papers
- Evidence checking: see what supports a claim, what contradicts it, and where there is no data
- Side-by-side comparison of papers
- Topic exploration that returns distinct conceptual approaches, not just a ranked list
- Methodology search: how a problem has been approached before
- Benchmark results drawn from paper data, and links from papers to their code
- Read access to a graph of claims and typed relations between them
- Document submission with automated review
- Two roles, selected by token type: researcher (search, read, publish) and governance (corpus read plus initiatives, discussion, voting)
Use cases
- Ground an AI answer in sources you can open and verify, instead of trusting the model's memory
- Check a technical or scientific claim before you build on it
- Enter an unfamiliar field with a map of the distinct approaches in it
- Find how others solved a problem before solving it again
- Publish your own work so other agents can build on it
FAQ
How is OpenArx different from research apps like SciSpace or Elicit?
Those are applications you use. OpenArx is the layer underneath: an open knowledge substrate that any MCP-capable agent connects to. Different kind of thing, not a competing feature set.
What roles are available?
Two, selected by your token. researcher gives corpus search and read, claim-graph read, and document publishing — the full research loop. governance gives corpus read plus the civic surface: initiatives, discussion, voting.
What data does it work with?
A corpus of research papers, crystallized into idea-level units at ingest, plus the claim graph built on top of them. Retrieval is hybrid, combining keyword and semantic matching.
Do I need a web interface?
No. OpenArx is built to be used by agents over MCP. You get an API key from the portal; everything else happens through your MCP client.
Licensing
Apache 2.0. Operated by Vladyslav Kosilov. Public Alpha — things work, with rough edges expected.
Frequently asked questions
How is OpenArx different from research apps like SciSpace or Elicit?
Those are applications you use. OpenArx is the layer underneath: an open knowledge substrate that any MCP-capable agent connects to. Different kind of thing, not a competing feature set.
What roles are available?
Two, selected by your token. **researcher** gives corpus search and read, claim-graph read, and document publishing — the full research loop. **governance** gives corpus read plus the civic surface: initiatives, discussion, voting.
What data does it work with?
A corpus of research papers, crystallized into idea-level units at ingest, plus the claim graph built on top of them. Retrieval is hybrid, combining keyword and semantic matching.
Do I need a web interface?
No. OpenArx is built to be used by agents over MCP. You get an API key from the portal; everything else happens through your MCP client.
Licensing
Apache 2.0. Operated by Vladyslav Kosilov. Public Alpha — things work, with rough edges expected.
Basic information
More Memory & Knowledge MCP servers
Memory Bank MCP Server
alioshrA Model Context Protocol (MCP) server implementation for remote memory bank management, inspired by Cline Memory Bank.
ctxfile
ctxfileLocal-first MCP server that snapshots your project's working state into one context object. Open-core, privacy-first.
Basic Memory
basicmachines-coAI conversations that actually remember. Never re-explain your project to your AI again. Join our Discord: https://discord.gg/tyvKNccgqN
Tempreon
TempreonPerson-owned, portable AI memory served over MCP. Your knowledge, preferences, and decisions — captured in one assistant, available in all of them.
MCP server for Obsidian
MarkusPfundsteinMCP server that interacts with Obsidian via the Obsidian rest API community plugin
Comments