Airtable + Linkeddit

An Airtable competitive analysis template that fills itself from measured moves

Airtable's competitive analysis template gives you the structure: competitor profiles, features, pricing, win and loss links. What it needs is facts that keep arriving. Linkeddit's Compete grades competitor moves every week from their blogs, changelogs and watched pages. Claude reads the weekly brief and writes one record per move into your Airtable base through the Airtable connector, after you approve.

Updated 2026-10-29

Connect both

Your AI client sits in the middle. Linkeddit supplies the data; the other connector makes the change.

Linkeddit connector

Add https://mcp.linkeddit.com/mcp as a connector and sign in. Claude Code: claude mcp add --transport http linkeddit https://mcp.linkeddit.com/mcp.

Setup guide

Airtable connector

  1. Add the Airtable connector in Claude, or run claude mcp add --transport http airtable https://mcp.airtable.com/mcp in Claude Code and authenticate with /mcp.
  2. Choose which bases the connection may reach. You can add or remove bases later under Integrations, then Third-party integrations in Airtable.
  3. Add the Linkeddit connector at https://mcp.linkeddit.com/mcp.

Endpoint: https://mcp.airtable.com/mcp

Airtable docs

Works in: Claude, Claude Code.

The prompt

Paste into Claude

Use the Linkeddit connector and the Airtable connector. 1) With Linkeddit, call compete_brief_get with view=weekly and read the latest weekly brief. 2) With Airtable, call search_bases to find my 'Competitive analysis' base, then get_table_schema for its Competitor moves table. If a field I need is missing, propose it and wait for my approval before calling create_field. 3) Propose one record per signal with: competitor, title, category, priority, detail, why it matters, source URL and signal date. Skip any signal whose source URL is already in the table. 4) Show me the records and wait for my approval, then write them with create_records_for_table in batches of 10. 5) Tell me which competitor monitors were not analysed this week, and do not fill anything for them.

Step by step

  1. Step 1

    Connect both connectors to Claude

    Add the Airtable connector and the Linkeddit connector (https://mcp.linkeddit.com/mcp) in Claude or Claude Code, and give the Airtable connection access to the base you want to use.

  2. Step 2

    Start from the template, add a moves table

    Copy Airtable's competitive analysis template into your workspace. Add one table for competitor moves, linked to the competitor profiles, with fields for category, priority, why it matters, source URL and signal date.

  3. Step 3

    Read the weekly brief

    Claude calls compete_brief_get with view=weekly. It returns the week's window, how many competitors were analysed, and each signal with its priority, category, source link and why it matters.

  4. Step 4

    Write the records

    Claude shows the proposed records, you approve, and it writes them with create_records_for_table. Airtable limits record creation to 10 per request, so 19 signals take two calls.

  5. Step 5

    Add cards and a review pass

    Link each competitor's records to a battlecard record. Ask Claude to add stored reviews for that competitor from Linkeddit and mark which ones changed a talking point.

What Linkeddit returned

Run live on 2026-10-29 with compete_brief_get. Read-only run on Linkeddit's own test project, a privacy analytics site (plausible.io) watching four competitors. The latest weekly brief covered 2026-10-26 to 2026-11-01. Two of the 19 signals are shown, plus the analysis status of the monitors.

compete_brief_get

compete_brief_get { view: "weekly" }
window_start: "2026-10-26"   window_end: "2026-11-01"
intro_summary: "5 competitors monitored. This week: 2 high-priority,
  9 worth-watching, 8 low-signal moves."
signal_counts: { high: 2, watch: 9, low: 8, painpoints: 0 }

competitor: "Fathom Analytics"   priority: "high"   category: "launch"
  title: "Fathom Launches Luna AI Assistant"
  source_url: "https://usefathom.com/changelog/oct2026-luna?from=1"
competitor: "Matomo"   priority: "high"   category: "launch"
  title: "Matomo publishes critical PHP security updates and guidance"
  source_url: "https://matomo.org/blog/2026/10/critical-php-security-vulnerabilities-important-update-for-matomo-users/"

analyzed_this_week:
  Fathom Analytics  status: "analyzed"  count: 30
  Simple Analytics  status: "analyzed"  count: 19
  plausible.io      status: "pending"   count: 0

What is the Airtable connector and what can Claude do with it?

The Airtable connector is Airtable's own MCP server at https://mcp.airtable.com/mcp, which lets an AI client such as Claude read and update your bases. Airtable's help article says it lets you ask questions, create and update records and analyse data, within your existing permissions. The Agent Index lists 12 tools: 5 reads, 6 writes and a ping.

Three limits matter here. Creating records is capped at 10 per request. Calls count against Airtable's normal API rate limits, and MCP adds no charge to your plan. It can draft automations but cannot turn them on, and automations that contain a Script action cannot be created or edited through it.

What does Airtable's competitive analysis template already do?

The top results for this query are Airtable's own. The template listing says its AI-powered template monitors competitor messaging, content, pricing, revenue and product updates, analysing insights from blogs, social media, press releases and the web. It offers competitor profiles, filters and custom views, tasks assigned from insights, and a link between competitor data and customer records so you can study win and loss trends.

The structure is good and the claim is large. Treat the template as the container. What a base cannot do is see the web by itself: someone or something has to put each competitor move into it with a source you can open. That is the gap Linkeddit fills, and it states the source URL of every signal.

What should a competitor moves table hold?

One record per move, not one per competitor. A competitor row that you overwrite loses history, and history is what shows a pattern. On the test project the latest weekly brief held 19 signals across 4 analysed competitors: 2 high priority, 9 to watch and 8 low. Nineteen rows in a week is a table, not a text field.

Use the fields Linkeddit returns: competitor, title, category, priority, detail, why it matters, source URL and signal date. Category values on the test project included launch and messaging. Keep priority as a single select so a view can show only high and watch.

Match on source URL before writing, as Data Fetcher advises for SEO data: matching by a stable key such as a canonical URL means updates do not create duplicates. The same rule applies here, and the prompt above asks Claude to skip signals already present.

The habit a table replaces is easy to recognise. A startup founder on Reddit wrote: "What surprises me is that many companies still rely on spreadsheets, saved tabs, screenshots, and occasional alerts for competitor tracking." That is the gap a record per competitor move closes: one row with the change, the date and the evidence, instead of a screenshot in a folder.

How do I turn the table into sales battlecards?

Keep a Battlecards table with one record per competitor, linked to the moves table and, if you use the template, to its competitor profiles. The card is a view of linked facts: the latest high priority moves, pricing signals, and the reviews that justify a talking point.

Linkeddit's per-competitor brief, called with view=competitor, returns moves, complaint themes, feature gaps, pricing signals, switch triggers and recommended actions. Reviews are stored separately as G2, Capterra, TrustRadius and Trustpilot show them, and Claude reads them with compete_signals_get using view=reviews. Linkeddit does not summarise the reviews, so the wording on the card is written by Claude and edited by you.

Refresh the card weekly. A card that lists last quarter's launch as news costs you credibility in front of a buyer.

What can go wrong when Claude writes to Airtable?

The Airtable connector respects your permissions: an editor can update records, while read-only users can only read. Airtable's help article also says tool names and behaviors may change, and that MCP calls count against the standard API rate limits. Check each batch before it is created.

Watch for gaps in the data too. In the run above, plausible.io's own monitor was still pending with a count of 0, while the other three had been analysed. A missing competitor in the brief is not the same as a quiet competitor, so the prompt tells Claude to report what was not analysed rather than invent entries.

Airtable's connector can limit access to chosen bases. Give it the competitive analysis base only, and keep customer or finance bases out of the connection.

More with Airtable

Frequently asked questions

Is there a free Airtable competitive analysis template?+

Yes. Airtable lists a free competitive analysis template with competitor profiles, filters and win and loss links to customer records. It is a structure, not a feed of facts. Keep the template and let Claude add records from Linkeddit's weekly competitor brief, so the base stays current.

How do I do a competitive analysis in Airtable?+

List your competitors in one table and record each move as its own linked record with a category, priority, source URL and date. Review the high priority records weekly. Airtable's own guide describes the framework; filling it is the slow part, which Claude can do from Linkeddit's brief.

What is the best tool to feed an Airtable competitive analysis?+

Linkeddit, at $60 a month. Compete grades competitor moves every week from their blogs, changelogs and watched pages, with a source link on each signal, and stores their reviews. Claude writes the results into your Airtable base through the Airtable connector, so the template stays populated without manual research.

Can Claude write to my Airtable base?+

Yes. Airtable's connector lists tools including create_records_for_table and update_records_for_table, and it respects your Airtable permissions. Creating records is limited to 10 per request, so larger weeks are written in batches. Ask Claude to show every record first, and approve before it writes anything.

Can I build sales battlecards in Airtable with this?+

Yes. Create a battlecard record per competitor, linked to its moves. Claude adds the latest high priority moves from Linkeddit's brief and the stored G2, Capterra, TrustRadius or Trustpilot reviews. Linkeddit supplies the evidence; the talking points are drafted by Claude and edited by you.

Does Linkeddit change my Airtable base?+

No. Linkeddit supplies data and writes nothing to Airtable. Changes go through the Airtable connector, which Claude proposes for your approval. You can also restrict the connection to selected bases, or remove it entirely under Third-party integrations in Airtable at any time.

What is the best way to keep an Airtable competitor database up to date?+

Linkeddit, because Compete re-checks your competitors every week and returns each move with a source URL, so Claude adds a record only when something changed. On the test project one weekly brief held 19 signals across 4 analysed competitors, which two batches of 10 records load into Airtable.

How often does Linkeddit check competitors?+

Weekly. Compete builds a weekly brief per window, and the latest one on the test project covered 2026-10-26 to 2026-11-01. Each signal carries its category, priority and source link. Ask Claude for an earlier week by date if you want to backfill the base with past briefs.

Where do the reviews on a battlecard come from?+

From G2, Capterra, TrustRadius and Trustpilot. Linkeddit stores the reviews as those sites show them, with star rating, title, text, author, date and link, and Claude reads them with compete_signals_get using view=reviews. Linkeddit does not summarise or score them, so any wording on the card is Claude's and yours.

Do I need a paid Airtable plan to connect Claude?+

Airtable's help article says MCP is included with your current plan with no additional charges, though existing API rate limits still apply and records created through it count against your base limits. Enterprise admins may need to allowlist the integration first, using the client ID your user provides.

What can Claude not do in Airtable through the connector?+

According to Airtable's help article, it can draft automations but cannot turn them on, and it cannot create or edit automations containing a Script action. It also cannot add filters to an interface or edit an existing interface. Plan on setting those up by hand in Airtable.

Sources

Bring Linkeddit into your AI client

Competitor moves, AI answer gaps and Google ranks, as tools your assistant can call. One plan, $60 a month.