MakerRadar
Search MakerRadar, a weekly product launch platform and startup directory: find products by keyword, platform, audience, use case or country, read one in full, list its alternatives, compare two side by side, browse categories, and follow this week's or any past week's launch board. Signed in, draft, edit and launch your own product listing. Remote (Streamable HTTP, no key for reads) at https://api.makerradar.io/mcp, or local via npx -y makerradar mcp serve.
Hosted MCP Server
npx add-mcp 'https://api.makerradar.io/mcp/account'Installs into Claude Code, Codex, Cursor and more
Documentation
Developers
MakerRadar MCP server
A Model Context Protocol server for MakerRadar's directory: AI assistants search launched products, read a listing, find alternatives and compare tools as function calls. Reading needs no key; acting for a person needs their approval.
On this page
- Connect
- Tools
- Try it
- Acting for a person
- How agents find it
Connect
The server is at https://api.makerradar.io/mcp — Streamable HTTP, JSON-RPC 2.0, stateless, no key. In Claude Code:
claude mcp add --transport http makerradar https://api.makerradar.io/mcp
In Claude.ai or ChatGPT (custom connectors), Cursor, VS Code or any client with a Connect button, add https://api.makerradar.io/mcp/account instead: the client signs in with OAuth and the person approves it on MakerRadar, so the account tools work too. Any other MCP client that supports remote HTTP servers takes https://api.makerradar.io/mcp as is.
Tools
Open to every agent, no key:
search_products(Search products) — Search MakerRadar's directory of launched products by keyword (matches name and tagline), ranked by all-time upvotes. Narrow it by category, tag, the technology it's built with, platform, audience, use case or the country it's based in (combined, a product matches them all); an unknown value is an error naming where the valid ones are listed. Omit everything to list the most-upvoted products.get_product(Get a product) — Get the full public details of one product by its slug.list_current_launches(This week's launches) — List the products on MakerRadar's live launch board (this ISO week's, or the last week until the new one opens), ranked by upvotes.top_products(Top products of all time) — List the most-upvoted products on MakerRadar of all time (the Top 100 board).list_categories(List categories) — List MakerRadar's categories, each with how many listed products it has and its page, most products first.get_alternatives(Alternatives to a product) — List the alternatives to one product (by its slug) in the order its alternatives page ranks them: editor's picks first, then makers' own declarations, then by relevance and upvotes.compare_products(Compare two products) — Compare two products (by their slugs): each one's public details side by side, which has more community upvotes, and the comparison page.get_launch_week(A launch week's results) — Get one launch week's full results (by its week, e.g. week-41-2026), ranked by upvotes — any week since MakerRadar began, not only the live one.list_launch_weeks(Every launch week) — List every launch week on MakerRadar, newest first: its dates, whether it's live, closed or upcoming, how many products launched and which won.list_tags(List tags) — List every tag on MakerRadar, each with how many listed products carry it and its page, most products first. Its slug is the `tag` filter of search_products.list_technologies(List technologies) — List every technology (framework, language, database, hosting…) listed products are built with, each with how many use it and its page. Its slug is the `builtWith` filter of search_products.list_facets(Platforms, audiences and use cases) — List MakerRadar's three facet vocabularies — platforms, audiences and use cases — each term with how many listed products it describes. Their slugs are the `platform`, `audience` and `use_case` filters of search_products.list_countries(List countries) — List every country a listed product is based in, with how many and its page, most products first. Its code is the `country` filter of search_products.list_collections(List collections) — List MakerRadar's editorial collections (hand-picked sets of products), each with its description, how many products it holds and its page.get_collection(A collection's products) — Get one editorial collection by its slug: its description and every product in it, in the editors' order.list_makers(The makers' leaderboard) — List the makers on MakerRadar ranked by their products' all-time upvotes, with how many products each has listed. Paged by cursor; keep going until nextCursor is null.get_maker(A maker's profile) — Get one maker's public profile by their handle — bio, links, when they joined — and every product they're a maker of, most upvoted first.list_reports(Monthly reports) — List the months MakerRadar has a launch report for, newest first, each with its launches, upvotes, page and CSV.get_report(A month's launch report) — Get one month's launch report (by month, e.g. 2026-10): launches, upvotes, comments, the median, how many start free, the category and pricing mix, and every launch ranked.list_honors(List honors) — List MakerRadar's honors — Products of the Week, Month and Year, editor's picks — each with what it means and how many products hold it.get_honor(An honor's products) — Get every product holding one honor (by its slug, e.g. products-of-the-week), newest first — the collection its page shows.list_glossary(The glossary) — List MakerRadar's glossary: every term (domain rating, dofollow, launch week…) with its one-sentence definition, longer explanation, related terms and page.list_posts(List blog posts) — List every published post on MakerRadar's blog, newest first: title, excerpt, author, whether it's sponsored, the product it reviews (if any) and its page. get_post carries the body.get_post(A blog post) — Get one published blog post by its slug, with its full body as Markdown.list_directories(Other launch directories) — List the other launch directories and product listing sites MakerRadar tracks for makers — each with its Domain Rating by Ahrefs, whether its link is dofollow, whether submitting is free and where to submit.check_page_seo(Check a page's SEO) — Check one web page's on-page SEO as MakerRadar's free SEO checker does: title and description length, one H1, canonical, language, indexing, Open Graph, structured data, image alt text and readable text — each check pass / improve / fix with how to fix it. It fetches that one page as our crawler (MakerRadarBot), after the site's robots.txt, and stores nothing; what it reads from the page is the site's own, unverified. Cached ten minutes; at most 20 checks an hour per network and 30 an hour per site checked.check_ai_visibility(Check AI visibility) — Check whether AI assistants and answer engines may read a site: its robots.txt rules for 20+ AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended…) and the search crawlers answers rely on, Content-Signals, an llms.txt, a Markdown version of the page, snippet rules, structured data and text readable without JavaScript. Fetches the page, robots.txt and llms.txt as our crawler after robots.txt; stores nothing. at most 20 checks an hour per network and 30 an hour per site checked.check_site_security(Check security headers) — Check a site's transport and headers: the plain-http → https redirect, HSTS, Content-Security-Policy (enforced or report-only, unsafe-inline), clickjacking protection, X-Content-Type-Options, Referrer-Policy, Permissions-Policy, mixed content and security.txt — each pass / improve / fix with how to fix it. Fetches the homepage once as our crawler after robots.txt, plus one request to the http:// address; stores nothing. at most 20 checks an hour per network and 30 an hour per site checked.check_email_dns(Check email DNS (SPF, DMARC, DKIM)) — Check a domain's email authentication from public DNS — SPF (present, single, within 10 lookups, how it ends), DMARC (policy, pct, reporting), a DKIM selector, BIMI, the mail and DNS providers, DNSSEC and CAA — each pass / improve / fix with how to fix it. DNS only: the site is never visited. at most 20 checks an hour per network and 30 an hour per site checked.lookup_domain(Look up a domain's registration) — Look up a domain's registration from the registry's RDAP record: registered, last changed and expiry dates, age in days, the registrar, status codes (transfer lock) and the registrant's country when published — never a person's name, address or email. Checks warn on a very new domain, a near expiry and a missing lock. No visit to the site. at most 20 checks an hour per network and 30 an hour per site checked.check_tls(Check the TLS certificate) — Check a host's TLS certificate from the public certificate-transparency logs: issuer, valid from/until, days left, wildcard, how many names it covers, and whether the zone's CAA records allow that issuer. No connection to the site. at most 20 checks an hour per network and 30 an hour per site checked.check_hosting(Find the host and CDN) — Find where a site is served from: its hosting platform (Vercel, Netlify, Cloudflare Pages, Shopify…), the CDN or edge in front of it and whether the origin is hidden, the web server, the CNAME chain and the origin network (ASN) and country. Reads DNS, the network registry and the homepage's headers (one fetch as our crawler, after robots.txt). at most 20 checks an hour per network and 30 an hour per site checked.detect_tech(Detect the tech stack) — Detect what a website is built with from its HTML, scripts and headers: framework or site builder, analytics and session replay, ad pixels, CRM and forms, support chat, payments, auth, consent banner, monitoring and fonts — each with the evidence that matched and a link to the products on MakerRadar built with it — plus whether the page is served by the server or rendered by scripts, and the third-party hosts it loads. One fetch as our crawler after robots.txt; stores nothing. at most 20 checks an hour per network and 30 an hour per site checked.preview_social_card(Preview a page's social card) — Show the card X, LinkedIn, Facebook and Slack display when a page is shared: its title, description, image (with the image's real size in pixels, read from the image itself), image description, X card type and og:url — each check pass / improve / fix with how to fix it. It fetches the page and the image its og:image names as our crawler (MakerRadarBot), after each host's robots.txt, and stores nothing; what it reads is the site's own, unverified. at most 20 checks an hour per network and 30 an hour per site checked.test_robots_txt(Test a URL against robots.txt) — Test whether a URL may be crawled under its site's robots.txt by Googlebot, Bingbot and every AI crawler MakerRadar tracks (GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot, Google-Extended …), with the group and the exact rule that decides each, plus the file's Sitemap lines, Content Signals, Crawl-delay and the lines crawlers skip. It fetches only robots.txt and stores nothing. at most 20 checks an hour per network and 30 an hour per site checked.inspect_structured_data(Inspect a page's structured data) — List every JSON-LD structured data node on a page (schema.org types such as Organization, SoftwareApplication, Product, FAQPage, BreadcrumbList, Article) with the fields Google's rich results need and the ones each node is missing, plus blocks that don't parse — each check pass / improve / fix with how to fix it. It fetches that one page as our crawler (MakerRadarBot), after the site's robots.txt, and stores nothing; what it reads is the site's own, unverified. at most 20 checks an hour per network and 30 an hour per site checked.read_page_as_markdown(Read a page as an AI assistant does) — Read a web page the way an AI assistant does: the HTML its server sends (no JavaScript), converted to Markdown by the same converter MakerRadar uses to answer `Accept: text/markdown`, with its heading outline, word count, whether the content only appears once JavaScript runs, and whether the page itself answers with Markdown. The Markdown is capped at 20,000 characters. It fetches that one page as our crawler (MakerRadarBot), after the site's robots.txt, and stores nothing; the text is the site's own, unverified. at most 20 checks an hour per network and 30 an hour per site checked.check_llms_txt(Check or draft a site's llms.txt) — Check a site's /llms.txt against the llms.txt proposal (an H1 title, a > summary, ## sections of [name](url) links), its size and whether /llms-full.txt exists — and, whatever it finds, a draft llms.txt built from the homepage's title, description and main links, ready to save. It fetches robots.txt, /llms.txt, /llms-full.txt and the homepage as our crawler (MakerRadarBot) and stores nothing. at most 20 checks an hour per network and 30 an hour per site checked.check_agent_readiness(Check a site's agent readiness) — Check which of the files and signals AI agents look for a site publishes: llms.txt and llms-full.txt, a Markdown answer to `Accept: text/markdown`, an MCP server card, an API catalog, an OpenAPI document, an agent skills index, Content Signals and AI crawler rules in robots.txt, and DNS-AID records — each check pass / improve / fix with how to fix it. Missing API files are improvements, not failures, for a site with no API. It requests each well-known file once as our crawler (MakerRadarBot), after robots.txt, and stores nothing. at most 20 checks an hour per network and 30 an hour per site checked.find_site_logo(Find a site's logo and favicons) — Find every icon a site declares — rel=icon, the Apple touch icon, mask icon, the web app manifest's icons, og:image and /favicon.ico — with each one's real size in pixels (read from the image), best first, the way MakerRadar picks a listing's logo. Icons are returned as addresses on the site; nothing is copied or stored. It fetches the page, its manifest and at most 8 images as our crawler (MakerRadarBot), after each host's robots.txt. at most 20 checks an hour per network and 30 an hour per site checked.check_sitemap(Check a site's sitemap) — Find a site's XML sitemap (its robots.txt Sitemap lines, else /sitemap.xml), read it — and a few of a sitemap index's sitemaps — and report the URL count, the newest lastmod, Google's 50,000-URL / 50 MB limits, invalid XML and URLs on another host or plain http, each check pass / improve / fix with how to fix it. It fetches robots.txt and the sitemaps as our crawler (MakerRadarBot) and stores nothing. at most 20 checks an hour per network and 30 an hour per site checked.generate_keyword_ideas(Suggest keywords for a product) — Suggest the searches people type to find a kind of product: from a few words about what it is (not its name — "invoicing software", "AI meeting notes"), buyer, comparison, how-to and question phrases, grouped by intent. Computed from patterns, with no search volumes and no outside service; nothing you send is stored. The same as the keyword generator page.check_links(Check a page for broken links) — Check every link on one web page: up to 50 of them (the site's own first, at most 10 on any other site), each asked once for its status — broken (4xx, 5xx), unreachable or redirecting — plus http:// links on an https page and links with no text or "click here" text, each check pass / improve / fix with how to fix it. It fetches the page and each link's headers as our crawler (MakerRadarBot), after each host's robots.txt, and stores nothing. at most 20 checks an hour per network and 30 an hour per site checked.check_ai_citation_readiness(Check a page's AI citation readiness) — Check how easy one page is for AI answer engines (ChatGPT search, Perplexity, Google's AI Overviews, Claude) to quote: a clear answer in its first 100 words, question-shaped headings and FAQ markup, numbers, sources and quotations, a named author, a last-updated date, lists and tables, a Markdown version, and whether the AI answer crawlers may fetch it — each check pass / improve / fix with how to fix it. A score is not a citation. It fetches that one page as our crawler (MakerRadarBot), after the site's robots.txt, and stores nothing. at most 20 checks an hour per network and 30 an hour per site checked.find_alternatives_by_url(Find alternatives to a product by URL) — Find the alternatives to a product from its website address (or its App Store, Google Play or GitHub page): the listing that address belongs to on MakerRadar and its listed alternatives, in its alternatives page's order (editors' picks, then its makers' declarations, then the rest), with what each is best for. Answers from our own catalogue — the site is never visited. A product that isn't listed yet has none; its maker can list it free.plan_directory_submissions(Plan where to submit a startup) — Plan where to submit a startup: the launch directories and listing sites MakerRadar tracks, filtered (free to submit, dofollow link, minimum Domain Rating, region) and ranked strongest first (Domain Rating by Ahrefs, then a followed link, then free), spaced out ten a week, with each one's submit address. Credit the Domain Rating as "Domain Rating by Ahrefs" wherever you show it.check_domain_availability(Check if a domain name is free) — Check whether a product name is free as a domain: the name (or a domain you give) on .com, .io, .app, .dev, .ai, .co, .so and .xyz, each looked up in its registry (RDAP) — free, taken (with when it was registered, when it expires and the registrar) or no answer. A free answer means the registry has no record; confirm at a registrar before relying on it. No site is visited. at most 20 checks an hour per network and 30 an hour per site checked.get_launch_checklist(Get the product launch checklist) — The product launch checklist: what a product's website needs before launch day (HTTPS, security headers, email DNS, working links, icons; titles, structured data, a sitemap, AI crawler access; a share image and keywords) and the launch-day and after-launch steps, in order, each with the free MakerRadar tool that checks it. Static; the same for everyone.grade_website(Grade a website) — Grade a website in one pass: one score and a letter, from four parts — found on Google (on-page SEO, structured data, sitemap), quoted by AI (AI crawler access, llms.txt, citation readiness), shared well (the social card) and safe to visit (HTTPS, security headers) — with the fixes that matter most first. It reads robots.txt and the page once, plus the sitemap, the share image and the security headers, as our crawler (MakerRadarBot), and stores nothing. at most 20 checks an hour per network and 30 an hour per site checked.check_redirects(Trace a URL's redirects) — Trace where a URL redirects, hop by hop: each hop's status code (301, 302, 303, 307, 308) and address, up to 10 hops, then where it ends (a page, an error or nothing) — with checks on the chain's length, loops, http:// to https://, permanent versus temporary redirects and a move to another site, each pass / improve / fix with how to fix it. It asks each address once as our crawler (MakerRadarBot), never reads a page's body, and stores nothing. at most 20 checks an hour per network and 30 an hour per site checked.preview_serp_snippet(Preview a page's Google result) — Preview how a page would show in Google's results: its title and description as shown, measured in pixels against Google's cut-off (about 600 for a title, 920 for a description) with where each is cut, the site name and breadcrumb line, and whether the page can be indexed at all — each check pass / improve / fix with how to fix it. It reads the page once as our crawler (MakerRadarBot), after its robots.txt, and stores nothing. at most 20 checks an hour per network and 30 an hour per site checked.generate_ai_robots_txt(Write robots.txt rules for AI crawlers) — Write robots.txt lines for AI crawlers: start from a preset (allow every AI crawler, block training only, or block them all), override any crawler by its user-agent token, add paths to keep every crawler out of, a sitemap and Content Signals (search, ai-input, ai-train). Answers the file's text, which crawlers it blocks and allows, and any input it ignored with why. Nothing is fetched or stored; Googlebot and Bingbot are never named, so search indexing is unchanged.get_pricing(Prices) — What a listing can buy on MakerRadar — Premium launch, featured and boosted weeks, ads, newsletter sponsorship, done-for-you submissions — each with its id, today's price and what it gets. Use before start_checkout. No sign-in needed.
Also open, and the only ones that store anything — a draft a person must sign in and submit before anything is published:
draft_listing(Draft a listing for a person to submit) — Draft a listing on MakerRadar for the person you're helping: give their product's website and, optionally, a name, tagline and description. Returns reviewUrl, a link the person opens to sign in, check the listing and submit it; nothing is published until they do (and MakerRadar reviews it). If the site is already listed, returns its page instead. No sign-in needed.connect_account(Connect to a person's account) — Ask the person you're helping to let you act for their MakerRadar account. Returns a short code and a link: show them both, they sign in and approve you on MakerRadar. Then call connect_account_finish with deviceCode until it returns your token. No sign-in needed to start.connect_account_finish(Finish connecting) — After connect_account: trade deviceCode for your agent token once the person has approved. Returns status pending (ask again after `interval` seconds), denied, expired, or approved with the token — shown once; send it as Authorization: Bearer <token> on every call.
For an agent acting for a person (they approve it first — see below):
check_domain_rating(Look up a domain's Domain Rating) — Look up a domain's Domain Rating (Ahrefs' 0–100 measure of its backlink profile) and its rank, from Ahrefs' free Domain Rating endpoint — credit it as "Domain Rating by Ahrefs" wherever you show it. Needs an agent token (connect an account first): each reading spends a shared daily budget. Cached; nothing else is read from the site.get_my_account(Your account) — Who you're acting for on MakerRadar: the account an agent token belongs to, whether it's ready to list products, and what the token may do. Needs an agent token (Authorization: Bearer mr_…).list_my_listings(Your listings) — The listings of the account you act for on MakerRadar, newest first: each one's status (draft, in review, live…), its page once public, and where its maker manages it. Needs an agent token with listings:read.submit_listing(Submit a listing) — Submit a product as a draft listing on MakerRadar, owned by the person you act for: its website and, optionally, a name, tagline and description (we fill in the rest from the site). Already listed sites come back as listed instead. Then get_my_listing; prove the person's ownership of the site (start_ownership_check → verify_ownership — a listing launches only once its site is proven) and choose_launch to send it for review. Needs listings:write.get_my_listing(One of your listings) — One listing of the person you act for, by id: everything you can edit (copy, category, pricing, price from, where it's based, best for, demo video, technologies, platforms, audiences, tags, links elsewhere, quick answers, logo and photos with their ids), its status, what's left to make it complete, and whether it can be sent for review. Needs listings:read.update_listing(Edit a listing) — Change any of a listing's details: name, tagline, description, category, pricingModel, priceFrom, basedIn, bestFor, demoVideo, technologies, platforms, audiences, tags, links. Send only what changes; null or an empty list clears a field. Categories, technologies, platforms, audiences and tags are slugs from get_listing_options. A demo video must be a YouTube, Vimeo, Loom or Arcade link to a real, public video. Links are the listing's pages elsewhere (X, GitHub, LinkedIn, App Store, Google Play, Product Hunt, YouTube channel, Discord, Instagram), one per kind. A draft changes at once; on a live listing a new name, category, video or link, or a rewrite, waits for review. Needs listings:write.get_launch_options(Launch weeks with room) — The launch weeks open to a listing: Free (with spots left per week) and Premium. Use before choose_launch. Needs launches:write.choose_launch(Choose a launch and send for review) — Choose Free or Premium (and optionally a week, as get_launch_options names it) and send the listing for review. The person's ownership of the site must be proven first (start_ownership_check → verify_ownership), else this answers owner_unverified. MakerRadar reviews every listing before it goes live. Needs launches:write.start_ownership_check(Start proving the site is theirs) — Start proving the person controls a listing's website: method dns_txt, html_file or meta_tag returns a token and exactly where to put it. If you can deploy the site or edit its DNS, place it, then call verify_ownership. A proven owner can't have the listing taken over. Needs listings:write.verify_ownership(Verify the site is theirs) — Check the token from start_ownership_check is in place; if it is, the person becomes the listing's verified owner. DNS can take a few minutes to show. Needs listings:write.add_listing_image(Add a logo or photo) — Add a listing's logo (kind logo, replaces the old one) or a photo (kind photo) from a public https image URL. We fetch and check it like an upload; on a live listing it waits for review. Needs listings:write.get_listing_options(What a listing's fields accept) — Everything update_listing accepts: the categories, technologies, platforms, audiences and tags to pick from (by slug), the pricing models, currencies and price periods, the demo video hosts, the link kinds with an example of each, and each field's limits. Call it before editing so you never guess a value. Needs listings:read.set_quick_answers(Set a listing's quick answers) — Replace a listing's quick answers (the questions buyers ask, answered by the maker): up to 6, each a question of 8–140 characters and an answer of 2–600. An empty list removes them. On a live listing, answers that add a link or are mostly rewritten wait for review. Needs listings:write.remove_listing_photo(Remove a photo) — Remove one photo from a listing's gallery, by its id from get_my_listing. It's gone at once. Needs listings:write.set_listing_cover(Choose the cover photo) — Make one of a listing's photos (by its id from get_my_listing) the cover its page leads with. Needs listings:write.get_badge_snippet(The badge for their site) — The HTML badge for the listing's own website: put it on the homepage and, while it's there, the listing earns what MakerRadar's badge earns (the answer says what). Needs listings:read.start_checkout(Start a checkout for them to pay) — Start a checkout for the person you act for: an item from get_pricing for one of their listings (a placement needs the listing approved first). Returns checkoutUrl, the payment page: give it to the person to pay — you never pay, and nothing is charged until they do. We also email them the link. Holds a week's spot while they pay. Needs billing:checkout.list_my_orders(Your orders) — The orders of the person you act for, newest first: what each bought, for which listing and week, and whether it's paid. Use after start_checkout to see when they've paid. Needs billing:checkout.
Try it
One call, straight from a terminal:
curl -s https://api.makerradar.io/mcp -H 'content-type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"search_products","arguments":{"query":"analytics","limit":5}}}'
Tool results put what makers wrote under makerProvided, labelled “maker-provided, unverified”, with links and other sites' addresses taken out — so an agent can tell a listing's claims from our facts.
Acting for a person
The account tools list and launch a person's own products and start checkouts for them to pay. They need an agent token (mr_…), scoped and revocable, which the person approves — never their session, and no token can vote, comment or review:
- A client with a Connect button signs in with OAuth at
https://api.makerradar.io/mcp/account. - Any other agent calls
connect_account, shows the person a short code to approve at https://makerradar.io/connect, then collects its token withconnect_account_finish. - Or the person makes a key in Dashboard › AI agents.
The steps, for agents: /auth.md.
How agents find it
- /.well-known/mcp/server-card.json — the server card, on this site and the API host, as
io.makerradar/mcp. - /.well-known/ai-catalog.json — the AI catalog that lists it beside the agent skills and the REST API.
- /.well-known/agent-skills/index.json — agent skills that say when and how to use it.
- The REST API covers the same directory: MakerRadar API docs.
This is the Markdown version of https://makerradar.io/mcp on MakerRadar. Every page answers Accept: text/markdown.