Skip to main content
All posts

Let your AI assistant look after your LinkLynx page

LinkLynx now has a remote MCP server. Connect Claude Code, Codex, Cursor or VS Code with one token, then ask for a review, a rewrite or a new row.

Rafael Valle4 min read

LinkLynx now has a remote MCP server. That means an AI assistant can look at your page and, if you allow it, change it.

You ask in plain words. "Review my page and suggest a stronger first row." "Tighten my card titles." "Schedule this link for Friday at 9." The assistant does the work through LinkLynx, and you stay in charge.

MCP is a standard way for AI tools to talk to other products. If your assistant supports remote MCP servers, it can connect to LinkLynx.

What you need

  • A page on Premium or Studio.
  • A personal access token. Open the editor, choose Features, then Developer API, and create one. Copy it right away, because it is shown once.
  • An MCP-compatible assistant.

The endpoint is:

text
https://www.linklynx.bio/api/mcp

It uses the same token and the same rules as the REST API. Your plan limits apply. Give the token the read scope to let the assistant look. Add write to let it change things. Revoke the token in the editor and the assistant stops at once.

Connect Claude Code

One command:

bash
claude mcp add --transport http linklynx https://www.linklynx.bio/api/mcp --header "Authorization: Bearer llk_..."

Replace llk_... with your token.

Connect Codex

Codex works today with a token. Put the token in an environment variable named LINKLYNX_TOKEN, then add this to your Codex config.toml:

toml
[mcp_servers.linklynx]
url = "https://www.linklynx.bio/api/mcp"
bearer_token_env_var = "LINKLYNX_TOKEN"

Connect Cursor

Cursor takes a url plus headers entry in its MCP config. In .cursor/mcp.json:

json
{
  "mcpServers": {
    "linklynx": {
      "url": "https://www.linklynx.bio/api/mcp",
      "headers": { "Authorization": "Bearer llk_..." }
    }
  }
}

Connect VS Code

In VS Code it is an http server with headers. In .vscode/mcp.json:

json
{
  "servers": {
    "linklynx": {
      "type": "http",
      "url": "https://www.linklynx.bio/api/mcp",
      "headers": { "Authorization": "Bearer llk_..." }
    }
  }
}

These config files often end up in git. Keep the token out of version control: add the file to .gitignore, or read the token from an environment variable (Cursor supports ${env:LINKLYNX_TOKEN} and VS Code supports ${env:LINKLYNX_TOKEN} or an inputs prompt). If a token ever leaks, revoke it in the editor and create a new one.

Config formats change from time to time. If your client looks different, the idea stays the same: the URL above, plus an Authorization header with Bearer and your token.

claude.ai and Claude Desktop

Open Settings, then Connectors, then add a custom connector. Enter the URL above and add a request header named Authorization with the value Bearer llk_..., using your token. Custom request headers on connectors are still rolling out, so if you do not see the option yet, use Claude Code or Claude Desktop with the steps above in the meantime.

What your assistant can do

Four tools only read.

  • get_page
  • list_cards
  • get_card
  • get_analytics

Nine tools change your page, and they need the write scope.

  • update_page
  • add_card
  • update_card
  • delete_card
  • reorder_cards
  • create_row
  • rename_row
  • reorder_rows
  • delete_row

Your assistant can also read linklynx://page, a plain outline of your rows and cards in order, so it knows the page before it edits. There is a redesign_my_page prompt too. It walks the assistant through a review and a plan, and it waits for your yes before it changes anything.

Video generation is not part of this yet.

Keeping it safe

I care about this part, so here is how it works.

  • The token is the control. A read token cannot change anything. A write token can only touch your own page.
  • Plan first, then edit. Your assistant is told to show you a plan before editing and to ask before any delete. The tools that delete are marked as destructive, so your client asks for approval too.
  • Everything is logged. Every change your assistant makes lands in the same audit log as API writes.
  • Limits still apply. Your plan limits hold, and cards you schedule stay hidden until their time.

One more thing. Only connect an assistant you trust. Anything it reads from the web can try to steer it, so skim what it plans to change before you say yes.

What about ChatGPT?

Coming soon. ChatGPT needs sign-in instead of a token, and that is not built yet. The same goes for adding LinkLynx from the claude.ai connector directory. For now, the token route above is how it works.

Try it

Start with a read token. Ask for a review of your page and see what it says. If you like the plan, make a write token and let it do the tidying.

The full setup, the tool list and the safety notes live on the developers page. If you want to call the API from your own code instead, start with this guide.

A good assistant, like a good lynx, looks before it leaps.

Rafael