API and MCP

Drive GenLink from your own scripts and AI clients over the API, or through the MCP server for Claude Desktop, Cursor, and Windsurf.

You'll haveYour own agent, script, or MCP client talking to your GenLink workspace.
Time10 minutes

You can drive GenLink from your own code or from an AI client. There is an HTTP API, and an MCP server that exposes GenLink as tools to Claude Desktop, Cursor, and Windsurf.

Note: the product is GenLink, but the developer surface still carries the old GenSend name in places: the npm package is gensend-mcp, the environment variable is GENSEND_API_TOKEN, and the tools are named gensend_*. Use those names exactly as written here.

Get a token#

The API and MCP both authenticate with a personal access token that acts as you, on one workspace.

  • In the app: Settings > API access is where a token is meant to be minted: "Connect your own agent, script, or MCP client to GenLink. Mint a personal access token below and send it as a Bearer header." A token is shown once, so copy it immediately, and you can revoke it later.

Warning: in-app token minting may not be available in your workspace yet. If the mint action does not return a token, use the MCP login helper below to get one, and skip the raw API until minting works for you.

  • From the MCP CLI: the reliable path today is a one-time password login that prints a token:
npx gensend-mcp login
# enter your email + password once; it prints a token.
# your password is never stored, only the token is used.

The HTTP API#

  • Base URL: https://backend.genlink.so/api
  • Auth header: Authorization: Bearer <token>
  • Rate limit: 300 requests a minute per token.

The main entry point is POST /manager/messages. You send a plain-English instruction and the workspace manager agent carries it out, the same agent you chat with in the app.

curl -X POST https://backend.genlink.so/api/manager/messages \
  -H "Authorization: Bearer $GENSEND_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"message":"Find 25 DR-60+ marketing blogs relevant to my AI tool and draft outreach"}'

Because the manager acts on your behalf, an instruction that sends does real outreach. Only send instructions you have reviewed.

The MCP server#

MCP (the Model Context Protocol) lets an AI client use GenLink as a set of tools. Run the server with your token:

GENSEND_API_TOKEN=<token> npx gensend-mcp

To wire it into Claude Desktop, Cursor, or Windsurf, add it to that client's MCP config:

{
  "mcpServers": {
    "gensend": {
      "command": "npx",
      "args": ["gensend-mcp"],
      "env": { "GENSEND_API_TOKEN": "<token>" }
    }
  }
}

The tools#

  • gensend_whoami, confirm the token works and show the connected account and active workspace.
  • gensend_list_workspaces, list workspaces you can access.
  • gensend_select_workspace, choose which workspace later actions target.
  • gensend_list_campaigns, list campaigns with status, type, and counts.
  • gensend_get_campaign, one campaign in detail.
  • gensend_run_agent, the main tool: hand GenLink's operator agent a plain-English instruction.
  • gensend_launch_campaign, launch a campaign so it begins sending.
  • gensend_set_campaign_status, pause or resume a campaign.

Warning: reads are safe, but gensend_run_agent and gensend_launch_campaign can send real email. Only run them once you have approved sending, and remember the workspace approval gate and do-not-send list still apply.