Automate competitive intelligence in Make with Claude and Linkeddit

Make's own guide, published 14 October 2026, builds a sourced weekly brief. Its Make AI Web Search app has a Generate a response module that takes a prompt, such as a request to summarise a competitor's recent pricing changes, and returns a written answer plus the source pages it used. The brief is structured as competitor, signal, one-line summary and link, then posted through Slack on a Monday.

The guide tells readers to start with two or three competitors, because too many sources produce more noise than insight. It also says to treat every finding as provisional until it is sourced. Both are good rules, and the workflow on this page keeps them: it starts from the monitors you already track and sends a source link with every item.

Many teams start with something simpler than a scenario. "What surprises me is that many companies still rely on spreadsheets, saved tabs, screenshots, and occasional alerts for competitor tracking," said a startup founder on Reddit. A Make scenario that posts Linkeddit's dated moves replaces that pile with one feed.

A web search summary tells you what a competitor published. It does not tell you which of those moves matters, and it does not record what the answer engines say about you. Linkeddit's signals arrive already graded: on the test project each one carried a priority, a category such as messaging or other, a why it matters line and the source link.

Linkeddit also records where ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews recommend a competitor instead of you. Make's guide covers public web pages and does not cover that, so the two approaches stack: Make delivers, and Linkeddit decides what is worth delivering.

Claude finds the scenario with scenarios_list, reads its inputs with scenarios_interface, and runs it with scenarios_run, passing the message as an input. The Make connector lists 67 tools, 31 of them writes, and scenarios_run is the one that does the work here. Make's documentation describes inputs as parameters a client fills when the scenario runs.

Keep the scenario small. One text input and one posting module is easy to review, and it keeps the routing in Make, where you can see it. Do not let Claude create or edit scenarios for this job. The prompt above names the one scenario it may run.

No. Claude works when you start it, so you run the saved prompt each week or use a scheduler you already have to open a Claude session. Make's guide runs on a Make schedule, which is a different design. The reason this page starts from Claude is that Claude is the part that reads the Linkeddit connector.

Linkeddit's side does run on its own. On the test project all five competitor monitors are weekly, and each shows a next run of 2026-11-04. If you start the prompt the morning after a run, the data is fresh. If you need an alert the minute something changes, this is not that tool: it is a weekly, reviewed hand-off.

Check three things. First, that the filter is right: in the live run all five signals for the week of 2026-10-26 were low priority, and a prompt that posts every signal would have filled a channel with items nobody asked for. Second, that each item has a source link, so a reader can verify it.

Third, that a timeout is not a failure. Make's documentation gives a 25 second limit for scenario run tools over OAuth, and says a scenario that times out keeps running for up to 40 minutes, returning an executionId. Claude should read the result with executions_get before it runs the scenario a second time, or the post lands twice.