Jotform and Linkeddit in Claude: collect what customers say about your competitors, then test it

Jotform's connector lists 11 tools: 6 reads, 4 writes and 1 the listing does not classify. Jotform's developer page describes a default server with 7 tools. analyze_submissions, fetch, get_submissions and search are read-only. create_form and assign_form write data, and edit_form is marked destructive, because it can change or overwrite an existing form. The connector listing adds form_list, list_submissions, display_form and create_submission, which belong to the interactive MCP App version.

Jotform describes two servers that overlap on forms and submissions: the text-based MCP at https://mcp.jotform.com and the MCP App at https://mcp.jotform.com/mcp-app, which can show an interface in clients such as Claude. Its docs say you usually need only one of them. Both use OAuth sign-in, and Jotform API keys are not accepted.

The tool that matters here is analyze_submissions, described by Jotform as deep AI-powered analysis on one or more forms' submissions. That makes Jotform the better fit when the answers are sentences rather than counts.

Because a multiple-choice list can only return the reasons you thought of. A sales manager on Reddit named what call reviews miss: "But that doesn't tell you much about why a prospect ghosted after a demo or why one AE keeps closing deals while others struggle." A short open question after a purchase, a cancellation or a demo gets the reason in the customer's words.

Good prompts are narrow. 'What did you use before us, and what made you switch?' returns a competitor name and a reason. 'What nearly stopped you from buying?' returns an objection. Jotform's own description of the connector says Claude can filter responses with natural language, summarize feedback, compare results and uncover recurring themes, which is the reading job this survey needs.

If your survey is mostly multiple choice that you want counted, ranked and trended, the Typeform hub covers that style with its insights tools. Many teams will use both: one counted question, one open one.

Linkeddit gives Claude evidence that did not come from the same customers. Each competitor monitor in Compete lists the sources it watches, and Answer Radar lists the buying questions where the five answer engines (ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews) leave you out. Both come with one plan at $60 a month. Linkeddit does no theme analysis of its own. It returns data and the customer's own assistant reads it.

A read-only run on Linkeddit's test project, plausible.io, returned this from compete_monitors_get: 5 monitors, all weekly, next run 2026-11-04. Matomo, Umami, Simple Analytics and Fathom Analytics each watch blog_changelog, page_monitor, signals, g2, capterra and trustpilot. The Plausible Analytics monitor watches trustpilot. answer_gaps_get returned one open gap, rank 1, confidence 1: 'Not represented for: Umami vs Matomo for self-hosted analytics'.

So if survey themes from Jotform say customers care about self-hosting, Claude can point to a gap where an engine answers that exact comparison without naming you. If customers say a rival's support is better, the monitors show which review sites each rival is watched on, so the claim can be checked in the places buyers read.

First, ask Claude to call compete_monitors_get and answer_gaps_get and list the competitors and the open gaps. Second, ask it to draft a short form with create_form, with one required question naming the rival and one open question for the reason. Claude should not run edit_form on a live form. Ask it to build a new form so the destructive tool never touches one with responses in it.

Once answers come in, Claude uses get_submissions to list the submissions, fetch to open any single one, and analyze_submissions to group the themes. Ask it to quote three real answers per theme and give the submission count for each, so a theme backed by two people is not presented as a trend.

Last, Claude goes back to Linkeddit and sets each theme beside the data: does a rival's review source, a competitor signal or an answer gap match it? End with three columns: what customers said, what the data shows, and where they disagree.

Check the sample. A summary of eight answers is a handful of anecdotes however confident it reads. Check selection: people who answer a survey are not all your customers, and a survey sent only to happy accounts will not surface why others left. Check the connector's own warning on writes: edit_form is marked destructive, and Jotform's account settings let you view and revoke connected clients under My Account, then Connected Apps.

Check privacy. Jotform says HIPAA-protected data is only reachable on a HIPAA-enabled plan with the right scopes granted, and submissions can hold names and emails. Ask Claude to leave names out of its write-up. Check the data too: on the test project compete_brief_get returned an empty brief, so a good summary says what is missing instead of smoothing it over.