You can search public social media from the same chat you use to write code. Cursor and VS Code both connect to remote MCP servers, so an agent in your editor can find a relevant search, show you its price, run it once you approve, and put the source links next to your work. This guide covers the setup in each editor, the tools you get, three workflows that suit a code editor, and the precautions that matter when an agent that can edit files reads text written by strangers.
What you need before you start
- Cursor, or VS Code with GitHub Copilot Chat in agent mode.
- A Superlurk account. New accounts start with 10 free credits, and most searches cost 1 credit.
- An internet connection that can reach the published MCP host. Superlurk’s service is remote; your editor and any separately chosen bridge still run in their own environment.
- A question you want answered. The agent chooses searches better when you name the network, the topic and the time window.
Connect Cursor
- Use the current Cursor instructions. Open the Cursor integration guide for the server URL, one-click option and configuration instructions. Cursor’s official MCP documentation covers Customize and the supported configuration scopes.
- Review and authenticate. Review the endpoint and requested access, enable the server and complete the Superlurk browser sign-in. You can set a daily credit cap for this connection now or later.
- Verify one small task. Ask the agent to discover the Reddit post search and show its inputs and price before running it. Check that the returned sources refer to your intended question.
Connect VS Code
- Choose the intended harness and scope. Open the VS Code integration guide. The setup differs by agent harness and configuration destination; a portable file uses mcpServers, while the VS Code format uses servers. Do not copy one format into the other.
- Add the HTTP server through the guided flow. Use MCP: Add Server from the Command Palette, or the MCP Servers section in Agent Customizations, and enter the published server URL. Follow the prompts for the selected harness and destination.
- Authenticate and verify. Review the trust prompt, complete Superlurk browser sign-in and enable its tools in the intended agent chat. Organization policy can restrict servers. Use the official VS Code guide if the controls differ from your installed version.
Other editors and local-only clients
Other clients need compatible remote MCP transport and authentication support. Use the integrations hub for the supported setup route; a bridge for a local-only client is a separate software dependency, not a reason to run an arbitrary install command from a post. Client and organization policies can change which connections are available.
The tools your agent gets
| What it does | Spends credits | |
|---|---|---|
| list_networks | The networks you can search and how many searches each has. | No |
| find_searches | Finds the right search by words, network or what you want to find. | No |
| get_price | The exact price of a search, or of up to 50 accounts, links or IDs at once. | No |
| run_search | Runs a priced search and returns the first page of results. | Yes |
| get_run | Checks a search that is still running and pages through a bulk lookup. | No |
| list_results | Lists your saved results, newest first. | No |
| open_result | Opens a saved result again. | No |
| get_balance | Your credits, plan and this connection's daily cap. | No |
Only run_search spends Superlurk credits. Review its current quote before it runs, and check the editor’s actual tool-approval settings. Approval behavior depends on client permissions, organization policy and any existing authorization; do not assume every installation displays the same confirmation prompt. Use a connection-level daily credit cap when you need a service-enforced spending bound.
Three workflows that fit a code editor
Triage feedback after a release
After you ship, the useful feedback is scattered across X, Reddit and community threads. Ask the agent to gather it into a file in the repository: "Search X posts with Latest and Reddit posts for our product name since Tuesday's release. Show the prices first. Then open the comment threads of the three most discussed Reddit posts and write feedback.md, grouping complaints by feature with a link to each source." You approve each search as it comes up, and the result is a reviewable file instead of a chat summary.
See how developers describe an error
An exact error message in quotes is a precise query. Ask the agent to search Reddit posts for it, read the most relevant threads and list the fixes people report, each with its link. Check a fix against your own code and the library's documentation before applying it; a highly voted comment can still be out of date.
Collect public examples for docs or a demo
When you write a tutorial or plan a demo, it helps to see how others explain the same thing. Ask for YouTube videos about your library or category, then note the titles, channels and dates in a table. Product Hunt launch pages, which Superlurk finds on the web, show how similar tools describe themselves. Use these as references to write your own material, not as text to copy.
Treat search results as untrusted input
The same caution applies to links and exported files. Inspect destinations when needed for verification, and never execute retrieved text as a command. See social MCP security and the structured evidence export workflow.
Costs and limits
- Most searches cost 1 credit; deeper ones cost 5 or 25. The agent can show the exact price with
get_pricebefore anything runs, and a search that fails is free. - A daily cap per connection stops an editor from spending more than you intend. Change it or remove it in Superlurk under Settings, AI connections.
- An eligible bulk lookup can quote up to 50 accounts, links or IDs together. Review the valid item count, skipped items and total; each valid item remains its own retrieval.
- Searches return public content only, and a result is a sample of what the network served at that moment, not a complete archive.
- Every search the agent runs is saved to your Superlurk Results. Reopening a saved result and reading its saved pages are free. A fresh retrieval, refresh or additional upstream result page needs its own quote.
For a fuller picture of quotes, caps and bulk pricing, see social media MCP costs and limits.
Troubleshooting
- No Superlurk tools appear. Check the connection in the editor’s MCP controls and make sure it belongs to the harness or workspace used by the current chat. Complete browser sign-in, then enable the tools.
- The browser never opened. Check for a waiting sign-in tab, then use the editor’s connection or authentication control again. Follow its current client guide if the controls differ.
- Searches stopped with an authorization error. The connection was revoked or expired. Connect again from the editor.
- The agent says the daily cap is reached. Ask it to call
get_balanceto see the cap, then raise it under AI connections if you meant to spend more. - Nothing connects on a work network. A proxy may block
mcp.superlurk.com; ask whoever manages the network to allow it.
The social media MCP troubleshooting checklist covers more cases, including results that look empty or incomplete.
Disconnect when you are done
Disable or remove the server through your editor’s MCP controls or the configuration destination you used. To revoke the service authorization as well, use Settings › AI connections in Superlurk. Removing a configuration from one workspace does not necessarily remove a global or separately installed connection.
Frequently asked questions
Can the agent post, like or send messages on social networks?
No. The Superlurk tools search public content and read your own saved results. None of them publishes, reacts or messages anyone.
Where do the search results go?
They come back into the chat, so your editor's model processes them under the editor's terms. Each search is also saved to your Superlurk Results, so you can reopen it later without asking the agent again.
Do I need an API key?
No. You sign in to Superlurk in the browser the first time you connect, so there is nothing to copy into a config file or commit by accident.
Sources
How this guide was written. Reviewed on 11 November 2026 against Superlurk’s current integration guides and the official Cursor and VS Code MCP documentation. Setup links are the authority for current configuration; the workflows are illustrative, not measured outcomes. Approval prompts and organization policies vary by client and version.