Skip to main content
Connect AI assistants such as Claude, ChatGPT, VS Code, and Codex to Tamarind with the Model Context Protocol (MCP). Once connected, you can ask your assistant to:
  • Submit jobs for any Tamarind tool, including structure prediction, protein design, and docking
  • Check job status and get results
  • Upload and manage input files
For example: “Submit an RFdiffusion job to design a binder for this target protein” or “Check the status of my recent structure prediction jobs.” This page walks you through setup. For the latest credentials, client instructions, and the full list of MCP tools, see the MCP Server page in the app.

1. Open the MCP Server page

Sign in to app.tamarind.bio, click API in the left sidebar, then click MCP Server (1) under Setup. The page summarizes what the MCP server can do (2) and has everything you need to connect. MCP Server page in the Tamarind app

2. Check the requirements

You need (1):
  • A Tamarind account — sign up if you don’t have one.
  • OAuth credentials or an API key.
  • An MCP-compatible client, such as Claude.ai, Claude Code, Claude Desktop, VS Code, ChatGPT, or Codex.
Requirements and OAuth credentials

3. Choose how to authenticate

OAuth gives each person their own access and is the best choice for most users. The OAuth 2.1 box (2) shows:
  • Client ID
  • Client Secret — hidden by default. Click the eye icon to show it, or the copy icon to copy it.
  • Authorization URL — https://clerk.tamarind.bio/oauth/authorize
  • Token URL — https://clerk.tamarind.bio/oauth/token
Copy the Client ID and Client Secret from this page when your client asks for them. Some clients (ChatGPT, Codex, and Claude Science) sign you in without asking for credentials.
Treat the Client Secret like a password. Don’t share it or commit it to code.

API key

An API key is simpler and works well for personal use or automation. Your client sends it in the x-api-key header. Get your key from the API Key page. Redirect URI request, API key, and Server URL
  1. Request a redirect URI — only needed for enterprise gateways or custom MCP clients that get a redirect_uri mismatch error.
  2. API Key — how API key authentication works.
  3. Server URL — https://mcp.tamarind.bio/mcp, used by every client.

4. Connect your client

The Client Setup Instructions section (2) on the MCP Server page has step-by-step instructions for each client. Click a client to expand it. Client setup instructions in the app The steps for each client are summarized below.
The client screenshots below are illustrations that show where Tamarind’s values go. Menus and labels change as these apps update, so follow your client’s current interface.

Claude.ai (web)

No installation required. Works with free and paid Claude accounts.
  1. Go to claude.ai, click your profile icon, and open Settings.
  2. Go to Connectors and click Add custom connector.
  3. Enter the connector details (1):
    • Name: Tamarind
    • Remote MCP server URL: https://mcp.tamarind.bio/mcp
  4. Expand Advanced settings and enter your OAuth Client ID and OAuth Client Secret (2). Both are required.
  5. Click Add (3), then sign in with your Tamarind account to authorize the connection.
Illustration: adding Tamarind as a custom connector in Claude.ai The same steps are on the MCP Server page in the app: Claude.ai steps on the MCP Server page
Uploading files from Claude? Files over 4 MB are uploaded to mcp.tamarind.bio with a command that runs in Claude’s sandbox. To allow it, open Settings → Capabilities, turn on Allow network egress, and add mcp.tamarind.bio under Additional allowed domains. Connecting the server doesn’t need this. Only large uploads do.

Claude Desktop

Option A: Connectors (recommended). Use the same steps as Claude.ai: open Settings → Connectors, click Add → Add custom connector, enter the name, server URL, and OAuth credentials, then click Add and sign in. Tamarind then appears in your connectors list. Illustration: adding Tamarind in Claude Desktop Option B: Config file with an API key. Open your Claude Desktop config file:
  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json
Add the Tamarind server, replace YOUR_API_KEY with your key, then fully quit and restart Claude Desktop:
The config file can also use OAuth. See the MCP Server page for that example.

Claude Code

Add the server from your terminal, then restart Claude Code:

VS Code

VS Code connects to MCP servers through GitHub Copilot Chat.
  1. Add the following to .vscode/mcp.json in your workspace (1), or open the Command Palette (Cmd/Ctrl+Shift+P) and run MCP: Add Server.
  2. VS Code asks for your API key the first time you use the server.
  3. In Copilot Chat, type @tamarind to use Tamarind tools (2).
Illustration: Tamarind in VS Code For more, see the VS Code MCP documentation.

ChatGPT (web)

Tamarind Bio is published in ChatGPT’s plugin directory, so there’s nothing to configure.
  1. Go to chatgpt.com and open Plugins.
  2. Search for Tamarind Bio (1) and click Install (2).
  3. Sign in with your Tamarind account when prompted.
Illustration: installing Tamarind Bio in ChatGPT

Codex (OpenAI)

Desktop app (recommended). Open Plugins in the Codex sidebar, search for Tamarind Bio, click Install (1), and complete the sign-in. The plugin includes the MCP server and Tamarind skills. Add the server manually (2). Use this only if you added the Tamarind-Bio/tamarind-agent-plugin repository as a plugin marketplace:
  1. Open Settings → Plugins → MCPs → Add server.
  2. Set Name to tamarind, Type to Streamable HTTP, and URL to https://mcp.tamarind.bio/mcp.
  3. Under Headers, add x-api-key with your API key.
  4. Save, then start a new task so Codex finds the tools.
Illustration: installing or adding Tamarind in Codex Codex CLI. Add your key to your shell profile (~/.zshrc or ~/.bashrc):
Then add the server to ~/.codex/config.toml:

Other MCP clients

Any MCP-compatible client can connect to https://mcp.tamarind.bio/mcp. Most clients need three things:
  1. Transport: remote / HTTP (sometimes called Streamable HTTP). Not STDIO.
  2. URL: https://mcp.tamarind.bio/mcp
  3. Authentication: either an x-api-key header with your API key, or OAuth with the credentials from the MCP Server page.

Cursor

Add Tamarind to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json in a project (1). Set TAMARIND_API_KEY in your environment so the key isn’t saved in the file.
After saving, check that the server is on and its tools are listed under Settings → Tools & MCP (2). Illustration: Tamarind in Cursor

Windsurf

Add Tamarind to ~/.codeium/windsurf/mcp_config.json (on Windows, %USERPROFILE%\.codeium\windsurf\mcp_config.json). Windsurf uses serverUrl for remote servers:
Then refresh the server list in Windsurf’s MCP settings.

Antigravity (Google)

Antigravity connects with OAuth. Open Settings → Installed MCP Servers → Open MCP Config and add:
Save the file, click Authenticate next to tamarind, and sign in. Until you authorize, the server shows an Unauthorized error. This is expected.

Claude Science

Claude Science registers its own OAuth client, so there’s nothing to paste. Open Settings → Connectors, click Add connector → Remote URL, enter the name Tamarind and the server URL, click Add, then Connect and sign in.

Clients that only support local (STDIO) servers

If your client can only start local servers, use mcp-remote to bridge to Tamarind (requires Node.js):
For your client’s exact file location and format, see its MCP documentation. If you get stuck, email info@tamarind.bio.

Available tools

The Tamarind MCP server exposes 42 tools (1). Each one is marked as Read-only, Write, or Destructive (2), so your client can ask before running anything that changes your data. Available tools on the MCP Server page Expand any tool on the MCP Server page to see its parameters.
Ask your assistant to estimate a job first. estimateTime shows how long a job will take and how many weighted hours it will use, before you submit. See Compute Hours.

Troubleshooting

Still stuck? Email info@tamarind.bio.

Learn more

  • MCP Server page in the app — credentials, client instructions, and every MCP tool with its parameters.
  • API — use Tamarind from code, the CLI, or the Python SDK.
  • Manage Files — upload and organize input files.