MCP server
Connect Claude, Claude Code, Cursor or any other MCP client to TeselaGen, so your AI assistant can work with the sequences, designs and notebook entries in your workspace.
Setup
The server is remote and handles sign-in itself, so there is nothing to install. Every client opens a browser window to sign in the first time you connect, then refreshes its access on its own.
Requirements
- A TeselaGen account on the deployment whose data you want to reach.
- Your TeselaGen password, or an API Password. If you sign in to TeselaGen with Google or single sign-on, generate an API Password in TeselaGen under Settings, then API Password, and use that.
- An MCP client that supports remote servers: Claude, Claude Code, Cursor and most editors with MCP support.
Connector URL
https://mcp.teselagen.com/mcpA bare https://mcp.teselagen.com is rewritten to that path, so either form works. On a multi-tenant deployment, add ?instance= with the first label of your TeselaGen hostname to sign in to that deployment: for community.teselagen.com it is community. To start in a specific workspace, add ?lab= with the workspace ID. To use both, join them with &.
https://mcp.teselagen.com/mcp?instance=community
https://mcp.teselagen.com/mcp?instance=community&lab=<workspace id>TeselaGen shows the exact URL and a config snippet for your deployment under Settings, then MCP Access. Copy it from there if you are not sure which deployment you are on.
Claude
- In Claude, open Settings, then Connectors, then Add custom connector.
- Paste your connector URL (https://mcp.teselagen.com/mcp) and save.
- Sign in with your TeselaGen email and password in the browser window that opens, then approve access.
Claude Code
To connect TeselaGen to Claude Code, run this in a terminal:
claude mcp add --scope user --transport http teselagen "https://mcp.teselagen.com/mcp"Run /mcp inside Claude Code to start the browser sign-in, then claude mcp list to confirm the server is connected. Sign-in finishes on a localhost address ending in /callback. That is the standard loopback redirect for desktop apps, not a misconfiguration.
Cursor and other IDEs
Add the server as a remote MCP server in your editor's MCP configuration file. Cursor keeps one at ~/.cursor/mcp.json; other editors use the same entry in their own file.
{
"mcpServers": {
"teselagen": { "url": "https://mcp.teselagen.com/mcp" }
}
}Clients that ask for an explicit transport take the same entry with "type": "http" next to the URL.
Authentication
- Sign-in happens on a TeselaGen page in your browser the first time you connect. Your client never sees your password.
- Nothing to register. The server uses standard OAuth sign-in, and clients register themselves, so there is no client ID or secret to configure.
- Staying signed in: your client refreshes access on its own. If it asks you to sign in again, the same browser sign-in opens.
- To revoke access, disconnect the TeselaGen connector in your client.
How credentials, tokens and logs are stored and retained is covered in MCP Server Privacy.
Workspaces
Each connection works in one of your workspaces: the one in ?lab= on the connector URL if you set it, otherwise one your account belongs to.
- Ask your assistant which workspaces you have, or to switch to another one.
- A new conversation starts in the connection's workspace.
FAQ
Do I need to install anything?
No. The server is remote. Add the connector URL to your client and sign in with your TeselaGen account in the browser window that opens.
Which connector URL do I use for my company's deployment?
TeselaGen shows the exact URL and a config snippet for your deployment under Settings, then MCP Access. Copy it from there.
My client config uses an /sse URL. Does it still work?
Yes. Older configs that end in /sse keep connecting, so there is nothing to change.
Can I store protected health information?
No. The connector is not intended for protected health information, so keep PHI out of the notebook content you create or edit through it. Regulated workloads need a Business Associate Agreement first. See MCP Server Privacy for how data is handled.