What is OpenEvidence MCP?

OpenEvidence MCP is an unofficial open-source Model Context Protocol server that connects your OpenEvidence browser session to AI clients.

Users with their own OpenEvidence account can connect OpenEvidence to Codex, Claude, Cursor, Cline, Continue, and other MCP-compatible clients.

Which MCP clients are supported?

Any client that supports local stdio MCP servers may work. Maintained examples focus on Codex and Claude-style configuration; Cursor, Cline, Continue, and compatible IDE hosts can use the same command/args pattern.

OpenEvidence MCP | Unofficial MCP Server

Overview

OpenEvidence MCP is an unofficial Model Context Protocol server that connects OpenEvidence via your own browser session. It supports local auth status checks, history retrieval, article fetch by ID, and question execution with completion polling. It is not affiliated with OpenEvidence, does not provide medical advice, and does not bypass access controls.

Availability Note

OpenEvidence availability may depend on region, account eligibility, and OpenEvidence policy. Public materials in May 2026 indicate verified U.S. HCP/NPI-centered access and EU/U.K. unavailability. This project does not bypass regional, account, or policy restrictions.

What Is OpenEvidence?

OpenEvidence is a medical research product. This repository provides an unofficial MCP bridge for users who already have their own OpenEvidence account.

Tool Surface

oe_auth_status: Validate account session through /api/auth/me.
oe_history_list: Read previous conversations through /api/article/list.
oe_article_get: Fetch full article payload by article ID.
oe_ask: Create a new question and optionally wait for completion.

Tested / Target MCP Clients

OpenAI Codex / Codex CLITarget

Claude CodeTarget

Claude Desktop / Claude appTarget

CursorCompatible

ClineCompatible

ContinueCompatible

VS Code / GitHub Copilot MCP environmentsExperimental

Windsurf / Zed / Replit / Gemini-style hostsExperimental

Maintained examples focus on Codex and Claude-style local MCP configuration. Other MCP-compatible hosts may work when they support local stdio server command/args.

Common Workflows

Research-question follow-ups with source-backed context
Treatment literature comparisons and quick evidence checks
Long research threads where history and follow-up continuity matter
AI-agent workflows that need OpenEvidence account context

Install Targets & MCP Clients

macOS
Windows
Ubuntu / Linux
Codex CLI / Claude Desktop / Claude Code / OpenClaw / Cursor / Cline / Continue

SEO / GEO / Parser Readiness

Static semantic HTML, no JS framework dependency
Canonical URL + OpenGraph + JSON-LD metadata
robots.txt and sitemap.xml included
llms.txt and llms-full.txt for AI parsers
Dedicated semantic map: docs/SEMANTIC_CORE.md

Example Workflows

Research Follow-up: Ask OpenEvidence a research question, get source-backed answers with citations, then follow up in the same thread from your MCP client.
Treatment Comparison: Query two treatment approaches side by side. The MCP server preserves conversation context so follow-up questions refine the answer.
Journal Club Prep: Pull recent articles on a topic via oe_history_list and oe_article_get. Feed them into your AI workflow for summary and discussion points.
Team Knowledge Base: Multiple users can run their own local sessions. Each authenticates with their own OpenEvidence account - no shared credentials needed.