Connect Claude, ChatGPT, Cursor or your own agent to the Gensiv MCP server with OAuth sign-in or a workspace API key.
There are two ways to connect to https://app.gensiv.com/api/mcp:
| Method | Best for | Sees |
|---|---|---|
| OAuth | A person using Claude, ChatGPT or Cursor | Every workspace you belong to |
| API key | Agents, scripts and CI that run unattended | Every brand in the key's workspace |
Both are read-only by default. To let the assistant make changes, turn on Allow changes on the consent screen, or use a key created with Allow changes.
claude mcp add --transport http gensiv https://app.gensiv.com/api/mcpThen run /mcp in Claude Code and choose gensiv to sign in.
Add a custom connector (some apps call it a custom MCP server) and paste https://app.gensiv.com/api/mcp as the URL. The app opens Gensiv in your browser to sign in.
Add the server to the editor's MCP config:
{
"mcpServers": {
"gensiv": { "url": "https://app.gensiv.com/api/mcp" }
}
}Connecting uses OAuth, so you never paste a password into the assistant:
To disconnect, remove the connector in the app.
Create a key in Settings > API keys (owners and admins, Starter and up). The API reference explains keys, expiry and revoking. Send it in the x-api-key header, or as Authorization: Bearer gsk_....
claude mcp add --transport http gensiv https://app.gensiv.com/api/mcp \
--header "x-api-key: $GENSIV_API_KEY"Most MCP clients accept headers in their config:
{
"mcpServers": {
"gensiv": {
"type": "http",
"url": "https://app.gensiv.com/api/mcp",
"headers": { "x-api-key": "gsk_..." }
}
}
}A key skips the consent screen, so treat it like a password. It shares the API's limit of 120 requests a minute, and an invalid or revoked key gets 401.