Copy MCP, paste, run science
The Model Context Protocol lets assistants and coding agents call our scientific models as tools. Setup is three steps and takes under a minute once you have a key.
Setup steps
Request access
Use the Request Beta access form. Access is free during the Beta and reviewed by a person.
Receive your key by email
We email your API key manually. Keep it private and store it as an environment variable rather than in files you commit.
Paste the copied configuration
Pick your client below, click Copy MCP and paste into the location shown. Replace the placeholder with your key.
Your configuration
Snippets are generated for the current official configuration format of each client and include the endpoint plus an API-key placeholder. No real key is ever embedded.
Paste into: .mcp.json in your project (or ~/.claude.json for all projects)
- Export your key: NEBIUS_SCIENTIFIC_AI_API_KEY=... in your shell profile.
- Paste the JSON into .mcp.json at the root of your project.
- Run claude and approve the server when prompted, or use the one-line command below instead.
{
"mcpServers": {
"nebius-scientific-ai": {
"type": "http",
"url": "https://89.169.99.188/mcp",
"headers": {
"Authorization": "Bearer ${NEBIUS_SCIENTIFIC_AI_API_KEY}"
}
}
}
}Or add it with one command
claude mcp add --transport http nebius-scientific-ai https://89.169.99.188/mcp --header "Authorization: Bearer ${NEBIUS_SCIENTIFIC_AI_API_KEY}"Paste into: .cursor/mcp.json (project) or ~/.cursor/mcp.json (global)
- Export your key as NEBIUS_SCIENTIFIC_AI_API_KEY in your environment.
- Paste the JSON into .cursor/mcp.json.
- Enable the server under Cursor Settings → MCP.
{
"mcpServers": {
"nebius-scientific-ai": {
"url": "https://89.169.99.188/mcp",
"headers": {
"Authorization": "Bearer ${env:NEBIUS_SCIENTIFIC_AI_API_KEY}"
}
}
}
}Paste into: .vscode/mcp.json in your workspace
- Paste the JSON into .vscode/mcp.json.
- VS Code prompts for the API key input the first time the server starts and stores it securely.
- Start the server from the MCP view or when using agent mode.
{
"inputs": [
{
"type": "promptString",
"id": "nebius-scientific-ai-api-key",
"description": "Nebius Scientific AI Platform API key (sent to you by email)",
"password": true
}
],
"servers": {
"nebius-scientific-ai": {
"type": "http",
"url": "https://89.169.99.188/mcp",
"headers": {
"Authorization": "Bearer ${input:nebius-scientific-ai-api-key}"
}
}
}
}Paste into: ~/.codex/config.toml
- Export your key as NEBIUS_SCIENTIFIC_AI_API_KEY in your shell profile.
- Append the TOML block to ~/.codex/config.toml.
- Restart Codex; the server appears under /mcp.
[mcp_servers.nebius_scientific_ai] url = "https://89.169.99.188/mcp" bearer_token_env_var = "NEBIUS_SCIENTIFIC_AI_API_KEY"
Paste into: Settings → Connectors → Add custom connector
- Open Claude Desktop (or claude.ai) and go to Customize → Connectors.
- Choose Add custom connector and paste the endpoint URL below.
- Custom connectors authenticate through the connector flow; if you need header-based API-key access, use one of the developer clients instead.
https://89.169.99.188/mcp
Paste into: Any client that supports remote MCP servers over Streamable HTTP
- Use the endpoint URL as the MCP server URL.
- Send your key as an Authorization: Bearer header on every request.
- Consult your client documentation for the exact configuration file name.
{
"mcpServers": {
"nebius-scientific-ai": {
"transport": "streamable-http",
"url": "https://89.169.99.188/mcp",
"headers": {
"Authorization": "Bearer NEBIUS_SCIENTIFIC_AI_API_KEY"
}
}
}
}Replace NEBIUS_SCIENTIFIC_AI_API_KEY with the key you receive by email. Keys are never embedded on this page.
MCP questions
How do I receive my API key?
After your request is approved, we agree the available models, capacity and beta participation conditions with you. Keys are then created and emailed manually. We never publish keys on this site, and you should never paste a key into a public place.
Which models can I use right now?
The Models page polls our live catalog and marks each model as Available now, Preview or Coming soon. Access for your account, available capacity and applicable model licence terms are confirmed during onboarding. If the catalog cannot be reached the page says so instead of guessing.
What is MCP and why would I use it?
The Model Context Protocol lets AI assistants and coding agents call tools over a standard interface. Our MCP endpoint exposes the scientific models as tools, so you can ask an assistant to fold a sequence or dock a ligand without writing integration code.