Typeform and Linkeddit in Claude: survey your customers about competitors, then check the answers

The Typeform connector lists 60 tools: 24 reads, 30 writes, 1 admin action and 5 the listing does not classify. Typeform groups them by prefix. The forms tools create, edit and publish forms. The automations tools build follow-up steps with email, webhook, delay and integration actions. The contacts tools manage contacts and lists. The insights tools answer questions about responses: aggregate for counts, sums, averages and NPS, toplist for rankings, timeseries for change over time, and list for row-level data on a single field.

Typeform's own docs set the order of calls. Run accounts-list_accounts first, call forms-public_get_capabilities before building or editing a form, and call insights-public_discover before any analytics query, because field IDs are per form. Editing is a chain: get_capabilities, validate_patch, patch_form, then publish_form. patch_form saves a draft and does not publish it, so a survey stays private until you say it should go live.

One limit matters for survey work. Typeform's docs say bulk extraction of full response rows goes through its REST Responses API, not the MCP server. The MCP side is built for questions such as 'how many' and 'which first', so design the survey around answers you can count.

Because competitor pages tell you what a rival says, and only your customers can tell you who they weighed you against. A B2B SaaS founder on Reddit put the hard part plainly: "How do you get someone to entertain switching when they're not actively unhappy?" A short survey asks that question of people who already bought, or nearly did.

Typeform suits the counted version of this. A multiple-choice question such as 'Which other products did you consider?' becomes a ranking when Claude runs insights-public_toplist on it, and a time series with insights-public_timeseries shows whether a rival's name is appearing more often by week. A yes-or-no question about a feature you suspect a competitor leads on becomes a percentage with insights-public_aggregate.

Open-ended answers can be listed, but the connector is strongest on structured answers. If your survey is mostly free text, the Jotform hub covers that style.

Linkeddit checks the survey against the outside world. Compete tracks competitor blogs, changelogs, monitored pages and review sites and grades each move with a priority and a reason. Answer Radar records which brands the five answer engines name and recommend for the buying questions in your category. Both come with one plan at $60 a month. Linkeddit writes no survey and does no analysis of its own: the data is returned and your own assistant does the reading.

Here is a read-only run on Linkeddit's test project, plausible.io. compete_signals_get returned 19 signals for the week of 2026-10-26 across four tracked competitors. The excerpt: Fathom Analytics, priority high, category launch, 'Fathom Launches Luna AI Assistant', source Fathom Changelog, signal date 2026-10-01. Matomo, priority high, 'Matomo publishes critical PHP security updates and guidance', signal date 2026-10-01. Umami, priority watch, 'Introducing Umami MCP: Your Analytics, in Conversation'. The same project has five competitor monitors, all weekly, with the next run on 2026-11-04.

That is the check. If a survey of Plausible's customers says Fathom is the product they most often compared against, the signals show Fathom shipped an AI assistant and a new AI traffic tab in the same month.

First, Claude asks Linkeddit for the competitor list and recent signals with compete_monitors_get and compete_signals_get, and uses the tracked competitor names as the answer choices, so the survey lists the rivals that actually move. Second, it builds the form with the forms tools and stops at a draft. You read it and publish it.

After responses arrive, Claude calls insights-public_discover once, then aggregate, toplist and timeseries on the competitor question. It then returns to Linkeddit and lines up each competitor's share of mentions with that competitor's signals by week_of date. A rise that follows a launch is a lead worth testing. A rise with no matching move is a reason to ask a follow-up question.

Finally, ask Linkeddit for the answer gaps with answer_gaps_get. On the test project it returned one open gap, rank 1: 'Not represented for: Umami vs Matomo for self-hosted analytics'. A survey can then ask customers whether self-hosting was part of their decision, which tells you whether that gap matches how real buyers choose.

Check how many people answered. Typeform's insights tools count what they are given, and a handful of responses will not support a ranking. Check who the respondents are: customers who stayed will name competitors differently from people who left. Check that the question is closed and the choices are mutually clear, because insights-public_toplist can only rank the options you offered.

Treat any survey result as a claim about your customers, and Linkeddit data as a claim about the market. They answer different questions. When they agree, you have a stronger case. When they disagree, say so in the write-up instead of choosing the neater story. On the test project the competitor brief returned empty, and a good Claude summary reports that rather than filling the space.

Finally, keep personal data in mind. The contacts tools can store respondent details, and Typeform's connector requires you to grant all its scopes at once. Use an account you are comfortable letting Claude read, and ask only for what the analysis needs.