OpenEvidenceCAUTION
First open-source OpenEvidence MCP server: browser-session medical research tools for Codex, Claude Code, and MCP clients
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
No API key is required. Authenticate once with npm run login:session; the server reuses your local browser profile. Registry metadata version 0.2.2 continues to use the existing npm package 0.2.1.
OpenEvidence MCP is an unofficial Model Context Protocol server that connects OpenEvidence to Codex, Claude Code, Claude Desktop, Cursor, Cline, Continue, and other MCP-compatible clients through your own authenticated browser session.
[](https://www.apache.org/licenses/LICENSE-2.0) [](https://nodejs.org/) [](https://www.typescriptlang.org/) [](https://www.npmjs.com/package/@modelcontextprotocol/sdk) [](https://playwright.dev/)
[!IMPORTANT] This project is unofficial and is not affiliated with OpenEvidence. It does not provide medical advice, does not bypass access controls, and should only be used with your own OpenEvidence account in compliance with applicable terms, privacy rules, and clinical governance requirements.
Translations: Русский | Español | 简体中文 | 繁體中文(台灣) | 한국어 | हिन्दी
Agent Onboarding & Installation
Using Codex, Claude Code, Cursor, or another local AI coding agent? You can let the agent handle the entire setup, compilation, and local MCP configuration for you!
Copy and paste this short bootstrap prompt directly into your local AI coding assistant:
Please install OpenEvidence MCP for me: clone https://github.com/bakhtiersizhaev/openevidence-mcp, install dependencies, run build, auto-configure this MCP server in my local client (Claude Desktop/Codex/Cursor)
fbae72d72bf2OBSERVED · 2026-10-08Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add openevidence-mcp -- npx -y [email protected]
Exposed tools (5)
5 read · 0 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
oe_article_get | read | |
oe_article_wait | read | |
oe_ask | read | |
oe_auth_status | read | |
oe_history_list | read |
Trust audit
CAUTIONgrade C · trust 79/100 Install with care. The audit found things worth knowing before you trust its output.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | WARN |
| L2 | Instruction surface (what it tells the agent) | FAIL |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- none-observed
- Network
- declared (2 observation(s))
- Shell
- declared (5 observation(s))
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (7)
- Preserve unofficial, no-medical-advice, no-bypass, and user-session safety wording.
return await chromium.connectOverCDP(`http://127.0.0.1:${port}`);@modelcontextprotocol/sdk, dotenv, playwright, smol-toml, zod, @types/node, tsx, typescript
- It does not collect credentials.
This project uses the user's own authenticated browser session. The project does not collect credentials and does not require an official OpenEvidence API token.
- does not collect credentials;
This project is unofficial and is not affiliated with OpenEvidence. It does not provide medical advice, does not bypass authentication or access controls, and does not collect credentials. Users are r
Gates applied: instruction_override, no_behavioural_pass.
fbae72d72bf2full audit observations/trust-audit/mcp-server/bakhtiersizhaev__openevidence.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-08 | fbae72d72bf2 | CAUTION | C | 79 | first audit |
Questions
What is the OpenEvidence MCP server?
First open-source OpenEvidence MCP server: browser-session medical research tools for Codex, Claude Code, and MCP clients
What tools does OpenEvidence expose?
5 in total: 5 read-only, 0 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is OpenEvidence safe to connect to an agent?
With care. The audit graded it C (79/100) and found 7 things worth knowing before you trust this server, listed below with the exact line each was found on.
What credentials does OpenEvidence need?
It reads OE_MCP_AUTH_STATE_PATH from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does OpenEvidence run?
It speaks stdio, so it runs as a local process your client starts. It is published on npm as openevidence-mcp at 0.2.1.
How current is this page?
The grade is for one exact copy of the source (fbae72d72bf2), read on 2026-10-08. The repository is watched and re-audited when it changes.