Cloudflare MCP and Linkeddit in Claude: let AI crawlers in, find gaps, deploy fixes

The Agent Index lists 24 tools for the Cloudflare Developer Platform connector, 13 reads and 11 writes. They manage Workers (workers_list, workers_get_worker, workers_get_worker_code), D1 databases, KV namespaces, R2 buckets and Hyperdrive configs, plus accounts_list and set_active_account. Its description is that it builds applications with compute, storage and AI.

It does not list tools for WAF rules, bot settings or Pages deploys. Cloudflare's Claude Code setup page describes a separate API server that it says covers its whole API in about 1,000 tokens, plus the Wrangler CLI for deploys. So the connector is a good way to read what your Workers do, and the crawler and deploy work runs through the dashboard and Wrangler.

Allow OAI-SearchBot in Cloudflare's AI Crawl Control, which Cloudflare says is available on all plans. OpenAI says OAI-SearchBot surfaces sites in ChatGPT search, and that GPTBot, the training crawler, is a separate setting. Cloudflare's September 2026 changes let you choose Search, Training and Agent independently, with Disallow AI Training as the option that keeps search crawling on.

The full walk-through, with the defaults for new and existing domains and the rules that can still block a crawler, is the guide on allowing the ChatGPT crawler on Cloudflare.

Through Claude Code. Cloudflare's Pages documentation lists Git, Direct Upload and command-line project creation, and npx wrangler pages deploy followed by the build output directory is the documented Direct Upload command. Cloudflare's own Pages page also says Workers supports most Pages use cases and recommends starting new projects with Workers.

Linkeddit supplies the keyword research and stores the article Claude writes with article_save. The guide on publishing to Cloudflare Pages from Claude has the prompt and the preview-first steps.

The need is a stable address without a deployment stack. A Claude user on Reddit asked: "what workflow do you use if you want a stable link, working assets, and a way to replace the file without learning a full deployment stack?" A connector that publishes on your approval is one answer to that.

Linkeddit's gsc_queries_get reads the queries buyers used to reach a project's site from the customer's own Search Console. It has a striking distance view for queries at positions 11 to 20 with real impressions, a view of top positions that are clicked less than expected, and a view of queries with no dedicated page. The project must have Search Console connected.

Claude can take that list, find the matching pages in your repository, and propose a new title, a first-paragraph answer or a new section for each. Build, deploy a preview with Wrangler, check it, then promote it. Cloudflare's SEO guidance adds one check after a change: make sure your WAF custom rules, rate limiting and bot protection are not blocking Verified Bots.

Yes as a workflow, though it has little search demand of its own. Linkeddit's compete_signals_get returns competitor moves, snapshots of competitor pricing and changelog pages, and stored reviews with their ratings. search_competitors_get shows which competing domains rank for your keywords and how often the engines cite them.

Claude can draft a comparison from those facts and add it to your site as a page for review. Put in only claims you can source, such as a quoted price from a snapshot or a rating with its site. Deploy a preview and publish it yourself.

Two guides are live. The first covers allowing the ChatGPT crawler on a Cloudflare site and re-checking whether ChatGPT then names you. The second covers publishing an article from Claude to a Cloudflare Pages site, starting from a scored keyword and ending in a preview deploy that a person promotes.

The Search Console refresh and the comparison page above are covered in this hub rather than in separate guides.