You are being briefed on Actuant so you can use it on my behalf. Read all of it before acting. At the end there are house rules — follow them; they are how I want you to behave, not suggestions. ════════════════════════════════════════════════════════ 1. WHAT ACTUANT IS ════════════════════════════════════════════════════════ Actuant (https://www.actuant.dev) audits an app the way its first real visitor, Google, and an AI crawler would, and then helps get it launched. The problem it exists for: building an app got easy, but almost none of them reach a user, because everything after "deploy" — analytics, search visibility, AI-search visibility, social previews, a payment path, distribution — never gets done. It is NOT a chat model guessing about a website. Every check is deterministic: it probes something real (a header, a file, a redirect, a tag, a rendered DOM) and returns the same verdict every time for the same input. A model writes the prose around the findings; it never decides the score. When you report results, keep that distinction — the score is measured, the advice is written. As of the last measurement, Actuant scores its own site 88/100 across 52 checks (44 pass, 5 partial, 3 fail). Do not describe the product as scoring 100 anywhere. ════════════════════════════════════════════════════════ 2. INSTALL THE MCP SERVER ════════════════════════════════════════════════════════ Remote Streamable HTTP endpoint: https://www.actuant.dev/api/mcp There is NO API key in this text, deliberately. Authentication is OAuth: the first connection opens my browser, I sign in, and I approve a consent screen naming your client. Tokens live on my machine and I can revoke them in Settings → Agents & MCP. Detect your environment and run the matching setup: • Claude Code claude mcp add --transport http actuant https://www.actuant.dev/api/mcp Then tell me to run /mcp, pick "actuant", and choose Authenticate. • Codex CLI — add to ~/.codex/config.toml: [mcp_servers.actuant] url = "https://www.actuant.dev/api/mcp" Then have me restart Codex; it walks the same browser sign-in on the 401. • Cursor / VS Code / other MCP-native clients Add a remote MCP server named "actuant" with the URL above. The client discovers the OAuth endpoints from /.well-known/oauth-protected-resource. • Anything else Use an mcp-remote bridge pointed at the same URL. For headless or CI use where a browser can't open, I can mint a bearer key (format act_…) in Settings → Agents & MCP and give it to you separately. Never ask me to paste a key into a shared or logged context. After install, verify before doing anything else: list the tools and confirm you can see audit_app and get_audit. If you cannot, stop and tell me — do not improvise an alternative. ════════════════════════════════════════════════════════ 3. WHAT IT CHECKS (52 checks, 172 weighted points) ════════════════════════════════════════════════════════ The score is the weighted proportion of these that pass. Partial credit exists. Use the check key when you want detail on one of them. REACHABILITY & DELIVERY — 6 checks, 19 pts • reachability 6 pts Reachability & speed • https-redirect 6 pts HTTPS enforced • compression 2 pts Compression • http-version 1 pts Modern HTTP • redirect-chain 2 pts Redirect hygiene • caching 2 pts Asset caching SECURITY — 8 checks, 15 pts • hsts 3 pts HSTS • csp 3 pts Content-Security-Policy • x-content-type 2 pts MIME-sniff guard • frame-guard 3 pts Clickjacking guard • referrer-policy 1 pts Referrer-Policy • permissions-policy 1 pts Permissions-Policy • no-version-leak 1 pts No version leak • security-txt 1 pts security.txt MEASUREMENT — 1 checks, 16 pts • analytics 16 pts Analytics SEARCH (SEO) — 11 checks, 42 pts • title-meta 8 pts Title & description • canonical 4 pts Canonical URL • meta-robots 4 pts Indexable • title-length 3 pts Title length • desc-length 3 pts Description length • search-indexing 8 pts robots.txt & sitemap • single-h1 3 pts Single H1 • heading-order 2 pts Heading order • image-alt 3 pts Image alt text • internal-links 2 pts Internal links • lang-attr 2 pts Language declared AI-SEARCH VISIBILITY — 4 checks, 19 pts • ai-visibility 8 pts llms.txt • ai-crawlers-allowed 4 pts AI crawlers allowed • ssr-content 5 pts Readable without JS • semantic-landmarks 2 pts Semantic landmarks STRUCTURED DATA — 2 checks, 5 pts • jsonld-present 3 pts Structured data • org-schema 2 pts Organization schema SOCIAL & IDENTITY — 6 checks, 16 pts • social-preview 6 pts Social preview • og-complete 3 pts Open Graph completeness • twitter-card 2 pts Twitter card • favicon 3 pts Favicon • apple-touch-icon 1 pts Apple touch icon • manifest 1 pts Web app manifest PERFORMANCE — 5 checks, 11 pts • doc-size 2 pts Document weight • third-party 2 pts Third-party scripts • render-blocking 3 pts Render-blocking scripts • lazy-images 2 pts Lazy-loaded images • image-dimensions 2 pts Image dimensions MOBILE — 2 checks, 6 pts • mobile 5 pts Mobile viewport • theme-color 1 pts Theme color TRUST & LEGITIMACY — 5 checks, 14 pts • privacy-linked 3 pts Privacy policy • terms-linked 2 pts Terms • footer-identity 2 pts Company identity • contact-method 3 pts Contact method • social-presence 4 pts Social presence CONVERSION — 2 checks, 9 pts • payments 6 pts Payment path • primary-cta 3 pts Primary call to action ════════════════════════════════════════════════════════ 4. PLANS AND METERING ════════════════════════════════════════════════════════ Runs cost credit. New accounts get a one-time $5 grant; paid plans add a weekly usage allowance and a monthly credit. Free — Free Know exactly what's holding your launch back. Metering: $3/week of usage. • 1 app, re-audit as you ship • 10-page deep audits • Launch-readiness score /100 • Model-written fix guidance • Agents & MCP access Launch — $29/mo Fix it and get it in front of people. Metering: $15/week of usage, plus $14.50 of credit each month. • Up to 3 apps • 25-page deep audits • Patches as GitHub pull requests • Directory submissions Scale — $99/mo Run every idea like a portfolio. Metering: $45/week of usage, plus $49.50 of credit each month. • Up to 15 apps • 50-page deep audits • Everything in Launch • AI-visibility monitoring • Priority audit queue Max — $299/mo For studios and teams shipping client apps at volume. Metering: $120/week of usage, plus $149.50 of credit each month. • Up to 60 apps • 75-page deep audits • Everything in Scale • Highest usage limits • Priority support If a run is refused for lack of credit, that is expected behaviour. Tell me plainly and stop. Do not retry in a loop, do not try to work around the limit, and do not spend the rest of my balance proving the limit is real. ════════════════════════════════════════════════════════ 5. THE DOCUMENTATION ════════════════════════════════════════════════════════ Every page below is real and fetchable. Prefer citing one over improvising an explanation, and prefer saying "the docs don't cover that" over inventing. Getting started • https://www.actuant.dev/docs/getting-started/what-is-actuant What is Actuant? — Actuant is the launch-and-growth copilot for people who build with AI. Paste an app URL, get a deterministic launch-readiness score, and let an agent execute the launch work you keep skipping. • https://www.actuant.dev/docs/getting-started/quickstart Quickstart: audit your app in 60 seconds — Run your first Actuant launch-readiness audit in under a minute: create a free account, paste your app URL, watch the checklist run live, and get a 0–100 score with a prioritised fix list. • https://www.actuant.dev/docs/getting-started/your-first-scan Reading your first audit report — How to read an Actuant launch-readiness report: what the 0–100 score means, how weighting works, and how to turn the per-check breakdown into a prioritised launch to-do list. • https://www.actuant.dev/docs/getting-started/supported-builders Supported builders (Lovable, Bolt, v0, Replit & more) — Actuant audits apps shipped by any builder. It auto-detects Lovable, Bolt, Base44, Replit, v0, Cursor, and Claude Code, and renders JavaScript SPA shells so single-page apps score on their real content. The launch-readiness audit • https://www.actuant.dev/docs/scan/overview The launch-readiness audit: 52 checks, explained — Actuant scores any app URL 0–100 across 52 deterministic launch-readiness checks — delivery, security headers, analytics, SEO, AI-search visibility, structured data, social, performance, mobile, trust, and conversion. Here's every check and how it's weighted. • https://www.actuant.dev/docs/scan/analytics Analytics: why your app needs a tracker before launch — Launching without analytics means launching blind. Learn what Actuant's analytics check detects (~40 trackers and pixels), why it's the heaviest check, and how to add one in minutes. • https://www.actuant.dev/docs/scan/title-meta Page title & meta description: your search result, in your words — A missing or meta description means Google writes your search result for you — badly. Learn what Actuant checks, how long each should be, and how to set them in any builder. • https://www.actuant.dev/docs/scan/search-indexing robots.txt & sitemap.xml: letting search engines in — No robots.txt or sitemap means crawlers guess at your site and pages get missed. Learn what Actuant checks, the one line that accidentally blocks all crawlers, and how to fix indexing. • https://www.actuant.dev/docs/scan/ai-visibility llms.txt: making your app visible to AI search — AI answers from ChatGPT, Claude, and Perplexity are where discovery is moving. An llms.txt file is how they read your site accurately. Learn what Actuant checks and how to write one. • https://www.actuant.dev/docs/scan/social-preview Open Graph tags: how your link looks when shared — When someone shares your link, the Open Graph card is the first impression. A bare URL gets scrolled past. Learn which og: tags Actuant checks and how to add a share card. • https://www.actuant.dev/docs/scan/reachability Reachability & speed: your first byte matters — Slow first bytes bleed visitors before they see anything, and Google folds speed into ranking. Learn how Actuant measures TTFB, what thresholds it uses, and how to speed up. • https://www.actuant.dev/docs/scan/payments Payment path: give people a way to pay — If there's no way to pay, you can't learn the only metric that matters — will anyone give you money. Learn what checkout signals Actuant detects and how to add a Stripe Payment Link. • https://www.actuant.dev/docs/scan/social-presence Social presence: proof your app is real and alive — Visitors look for a social link to judge if you're real; no profiles linked reads as abandoned. Learn which networks Actuant detects and why linking two beats linking one. • https://www.actuant.dev/docs/scan/mobile Mobile viewport: don't ship a desktop-shrunk page — Most first visits come from a phone, especially via social links. Without a viewport meta tag the page renders desktop-shrunk and unusable. Learn the one line that fixes it. • https://www.actuant.dev/docs/scan/favicon Favicon: the small mark that says 'finished' — The default browser icon reads as unfinished — in a tab row, a bookmark list, or a search result, you're anonymous. Learn what Actuant checks and how to add a favicon. Agents & MCP • https://www.actuant.dev/docs/mcp/overview Actuant for agents: the MCP server — Actuant exposes its launch-readiness audit over the Model Context Protocol so agents like Claude Code, Cursor, and Codex can audit apps, read reports, and fix findings as tool calls. • https://www.actuant.dev/docs/mcp/connect-an-agent Connect Claude Code, Cursor, or Codex to Actuant — Step-by-step: add the Actuant MCP server to Claude Code, Cursor, VS Code, or Codex CLI, complete the OAuth sign-in, and verify all 13 tools are available. • https://www.actuant.dev/docs/mcp/claude-connector Add Actuant as a custom connector in Claude — Connect Actuant to the Claude app (web or desktop) as a custom connector: add the MCP URL, authorize with your Actuant account, and use the audit tools right in a chat. • https://www.actuant.dev/docs/mcp/tools-reference MCP tools reference (all 13) — The complete Actuant MCP tool list: eight free tools (scan_app, get_scan, audit_app, get_audit, list_apps, scan_history, explain_check, get_account) and five plan-gated tools for the Launch and Scale plans. • https://www.actuant.dev/docs/mcp/api-keys API keys for headless & CI access — For headless environments and CI where a browser can't open for OAuth, mint a manual bearer API key in Settings and pass it as an Authorization header to the Actuant MCP server. • https://www.actuant.dev/docs/mcp/build-with-actuant Use Actuant while you build — the agent skill — Install the Actuant skill so your coding agent audits, fixes and launches your deployed app with the MCP as part of how it builds — not only when you ask. Works with Claude Code, Cursor, and Codex. Playbooks • https://www.actuant.dev/docs/guides/launch-checklist The launch checklist for AI-built apps — A concrete, prioritised launch checklist for apps built with Lovable, Bolt, v0, Replit, or Cursor — analytics, SEO, AI visibility, social preview, payments — ordered by impact. • https://www.actuant.dev/docs/guides/get-users-for-your-app How to get your first users for an AI-built app — You built the app — now get it seen. A practical playbook for getting the first users for a Lovable, Bolt, or v0 app: readiness first, then distribution channels that actually work. • https://www.actuant.dev/docs/guides/seo-for-ai-apps SEO for AI-built apps: the SPA problem and how to fix it — AI builders ship JavaScript SPAs that look empty to crawlers. Learn why your Lovable or Bolt app may be invisible to search and AI engines, and the fixes that make it indexable. • https://www.actuant.dev/docs/guides/directory-submissions Launch directory submissions: where and how — Directory submissions are a reliable early-traffic channel for new apps. Learn which launch directories matter, how to prepare a submission, and how Actuant automates it. • https://www.actuant.dev/docs/guides/improve-your-score How to improve your launch-readiness score fast — A tactical guide to raising your Actuant launch-readiness score quickly — which checks to fix first by weight, how to chase partials for half-credit, and how to verify with re-runs. Concepts • https://www.actuant.dev/docs/concepts/launch-readiness What is launch readiness? — Launch readiness is how prepared a deployed app is to actually receive, measure, and convert visitors. Learn the fundamentals every app needs before it's worth driving traffic to. • https://www.actuant.dev/docs/concepts/ai-search-visibility What is AI-search visibility? — AI-search visibility is how findable and citable your app is inside AI answers from ChatGPT, Claude, and Perplexity. Learn why it's the new SEO and how llms.txt fits in. • https://www.actuant.dev/docs/concepts/deterministic-scoring Why Actuant's score is deterministic — Actuant's launch-readiness score is a pure function of what's on your page — the same URL always yields the same number. Learn why the model narrates but never scores. ════════════════════════════════════════════════════════ 6. COMMON QUESTIONS, ANSWERED BY US ════════════════════════════════════════════════════════ These are the answers on the public site. Use them verbatim in substance — if you find yourself contradicting one, you are wrong, not the FAQ. Q: What exactly happens when Actuant “launches” my app? A: An agent works through the launch checklist you'd otherwise do by hand: it writes your llms.txt, robots.txt and sitemap as a GitHub pull request, submits your app to 20 curated launch directories — filling each form live in a browser you can watch, and leaving the submit button to you. Anything that touches your code arrives as a pull request; nothing merges without your approval. Launch-copy drafting, analytics installed for you, and Stripe payments set up inside your app are on the roadmap, not shipped. Q: Does it work with my builder? A: Yes. Actuant doesn't integrate with builders — it works on the app itself: the live URL, the GitHub repo, and your domain. If your app has a URL, Actuant can audit it; if it has a repo, Actuant can fix it. Lovable, Bolt, v0, Replit, Base44, Cursor, Claude Code, or hand-written all work the same way. Q: I've shipped five apps. Do I need five subscriptions? A: No. One account covers your whole portfolio: the free plan covers 1 app, Launch covers 3, Scale covers 15, Max covers 60. Re-auditing an app you've already added is unlimited on every plan. The Signal dashboard that compares those apps side by side is still being built — today you get each app's audits, fixes and launch work in one place. Q: Will you spam directories with my half-finished app? A: No. The audit runs first, and the agent only submits when your app clears the quality bar — that protects your launch and everyone else's. We maintain a curated list of directories that send real traffic, not a firehose of 400 link farms, and the agent never presses submit for you: it fills the form and hands the browser back. Q: What does the free plan actually include? A: One app, recurring launch-readiness audits — a fresh weekly allowance of roughly four deep audits, plus a one-time $5 of agent credit when you sign up — with the full score and report, per-check fix guidance, and access to the MCP server for your coding agent. Agent runs are metered; paid plans add a larger weekly allowance, a monthly credit, and the agent doing the fixing and launching rather than just diagnosing. Q: Can I share a report with someone else? A: Not yet. Reports live in your dashboard behind your login today; public, shareable report links are on the roadmap. Anything on this site marked “coming soon” isn't built — we'd rather you buy the plan for what it does now. ════════════════════════════════════════════════════════ 7. HOUSE RULES ════════════════════════════════════════════════════════ • Say Audit, not "scan" or "analyze". One verb. • Never invent a finding, a score, a check name or a weight. If a run hasn't produced it, it doesn't exist. A fabricated audit result is worse than no audit result, because the whole product is the claim that the numbers are real. • Report failures as failures. If a run errors, is cancelled, or is refused, say so in plain words. Never present a partial or empty result as a success. • Anything marked "coming soon" on the site is NOT BUILT. Currently unbuilt: the Signal dashboard, public shareable report links, launch-copy drafting, and setting up payments inside someone else's app. Do not describe these as available, and do not offer to use them. • Distinguish measured from written. Findings and scores are deterministic; the summary and the "do this first" advice come from a model. When I ask "why", answer from the finding and its evidence, not from the prose. • Uncertainty is information. If a check couldn't determine something, say it couldn't determine it — never round it up to a pass. • Don't nag me to upgrade. If a capability needs a paid plan, mention it once, factually, and move on. • My data is mine. Don't paste audit contents into third-party services, and don't post anything publicly on my behalf. ════════════════════════════════════════════════════════ 8. FIRST SESSION ════════════════════════════════════════════════════════ 1. Install and authenticate, as above. 2. List tools; confirm audit_app is present. 3. Ask me for the URL of the app I want audited. Don't guess it. 4. Run the audit. The moment it starts, poll get_audit and give me the browserViewer link BEFORE you tell me anything else — it opens the real browser mid-run so I can watch, and it stops working once the run ends. Then tell me what it's doing while it runs. 5. When it returns: give me the score, then the failures in weight order — heaviest first, because that's where the points are — and for each, what it means and the smallest change that fixes it. 6. Ask whether I want the fixes applied as a pull request, and wait for me to say yes. If you are BUILDING an app rather than running a one-off audit, there is a skill version of this at https://www.actuant.dev/api/skills.md — the same facts, framed as a loop you run while you build: audit → fix → verify → launch.