Answerfox
The open-source AI-SEO toolkit that lives in your codebase and ships fixes as code.
Answerfox audits your site for SEO, AEO (answer engines), and GEO (generative engines), explains every issue in plain English, and gives you the fix as code you commit. It runs in your terminal and your CI, not as another dashboard you have to log into.
# Try without installing:
pnpm dlx @answerfox/cli audit https://stripe.comQuickstart
Install once, then run any command:
pnpm add -g @answerfox/cli
# or: npm install -g @answerfox/cliInstalling gives you two commands: answerfox (the full name) and af (a shorter alias). Pick whichever feels right.
# Audit any URL
af audit https://your-site.com
# Read the full doc for any check, in your terminal
af explain A4
# Scaffold trust pages, sitemap, and robots in a Next.js project
af init
# Scaffold agent-readiness manifests (works in any framework)
af add agent-card
af add mcp-server-card api-catalog agent-permissions oauth-discovery
# Fail a CI build if the score drops below a threshold
af audit ${PREVIEW_URL} --ci --min-score 80If you'd rather not install, pnpm dlx runs the CLI for a one-shot. Note the URL must be on the SAME line:
pnpm dlx @answerfox/cli audit https://your-site.comaudit and explain run anywhere and need no setup. init and add are interactive (they prompt before writing files) and expect a Next.js App Router project.
What a real audit looks like
Real output from audit https://example.com (a bare page, so it scores low, which is the point, it catches what is missing):
Audit · https://example.com
Fetched 2026-06-05T20:05:54.806Z
Score: 20/100 (Critical)
6 pass · 20 fail · 6 warn · 1 skip
[CRITICAL] 6 issue(s)
A1 · <title> present and 30-60 chars long
Title is 14 chars: "Example Domain"
Fix: Adjust the <title> to 30-60 characters (currently 14).
A4 · Canonical URL declared as an absolute http(s) link
Fix: Add <link rel="canonical" href="..."> inside <head>. Use defineSeo()
from @answerfox/metadata to emit this automatically.
C1 · At least one <script type="application/ld+json"> with valid JSON
Fix: Add a JSON-LD block. Start with organization() from @answerfox/schemas.
...Every finding names the check, why it matters, and the exact fix, often pointing at the Answerfox package that emits it. Run answerfox explain <id> for the full rationale of any check.
What good looks like
The score is a page-level machine-readability number, not a brand authority score. Audit a SaaS marketing page, a docs site, a content blog, or an indie product homepage. Avoid auditing logged-in product gates (like linkedin.com, github.com, twitter.com) where the public URL is intentionally sparse; the score will under-report because those pages are not built for crawlers.
For reference, here is how the tool calibrates on sites you already know:
| URL | Score | Why |
|---|---|---|
stripe.com |
76 | Full marketing surface, three real fixable gaps. |
calendly.com |
74 | Same shape as Stripe, a couple of trust-page misses. |
examples/basic-nextjs (in this repo) |
79 | What the init command scaffolds out of the box. |
vercel.com |
68 | Solid, a few real misses. |
github.com |
52 | Logged-out product gate, intentionally sparse. |
linkedin.com |
50 | Logged-out product gate, intentionally sparse. |
example.com |
20 | Bare page, no SEO at all. |
If your marketing page scores in the 70s with a few fixes flagged, you are doing well. If it scores in the 50s and you are not a login wall, the report is telling you something real.
You can audit any URL
Point Answerfox at any URL that returns HTML. Homepage, subpage, blog post, docs page, a GitHub repo page, a YouTube channel, a Medium article, a Notion public page, a Substack newsletter. The CLI fetches the page and analyzes the HTML, whether that page is yours or someone else's.
What the score means depends on who built the HTML you're auditing:
| If you audit... | The score measures... |
|---|---|
| Your own site or blog (you control the HTML) | Your work. Actionable. |
A subpage of your site (e.g. /pricing, /blog/post) |
Your work on that page. Actionable. |
| A competitor's marketing page | Their work. Useful for benchmarking. |
A GitHub repo page (github.com/<you>/<repo>) |
GitHub's repo template. Informational only. |
| A YouTube channel page | YouTube's template. Informational. |
| A LinkedIn profile or Medium article | Their platform template. Informational. |
For actionable fixes, audit a URL whose HTML you actually control. For curiosity, competitive research, or platform diagnostics, audit anyone.
A heads-up: github.com/<owner>/<repo> currently scores around 44 of 100, and linkedin.com scores 50. Big platforms are not yet optimizing their templated pages for AI search readability. Indie devs who host on their own domain just got a structural advantage in the AI search era.
Why Answerfox
Indie developers and solo founders lose hours of fiddly SEO work per launch on things that should be automated. Existing tools each cover a slice:
| Drop-in fixes | Continuous audit | Plain-English teaching | In your codebase | |
|---|---|---|---|---|
next-seo |
partial | |||
| Lighthouse | partial | |||
| Ahrefs / Semrush | ||||
| Answerfox |
The wedge: drop-in code, automated audit, and plain-English explanations, in one MIT-licensed toolkit that lives where you work.
What's in the box
Seven packages, each does one thing, all composable.
| Package | Purpose |
|---|---|
@answerfox/core |
Branded URL types, errors, the Check<T> interface every audit check implements |
@answerfox/schemas |
Type-safe JSON-LD generators (Organization, WebSite, FAQPage, Article, Breadcrumb, Product, SoftwareApplication, HowTo) |
@answerfox/metadata |
defineSeo() composes title, description, canonical, OpenGraph, Twitter, and robots into a Next.js Metadata object |
@answerfox/sitemap |
buildSitemap() with priority and changeFrequency inferred from path patterns |
@answerfox/templates |
Five trust-page templates (About, Privacy, Terms, FAQ, Contact) the CLI installs |
@answerfox/audit |
Cheerio-backed audit engine, 39 of 56 checks shipping today |
@answerfox/cli |
The answerfox command: audit, explain, init, add |
The audit framework
Checks are scored 0-100 (single score plus a band: Critical, Weak, Average, Strong, Excellent) and grouped by category. The framework defines 56 checks; 39 are active today.
| Category | Active | Total |
|---|---|---|
| A · Meta & technical | 9 | 10 |
| B · Content & chunking | 6 | 11 |
| C · Structured data | 2 | 10 |
| D · E-E-A-T & authority | 6 | 12 |
| E · Off-site citations | 4 | 8 |
| F · OpenGraph & social | 6 | 7 |
| G · Agent Readiness (informational, see note) | 6 | 6 |
| Total | 39 | 56 |
The remaining 17 (content-quality NLP, page-type heuristics, multi-page crawls) land incrementally. Full spec in AUDIT-FRAMEWORK.md.
About Category G (Agent Readiness): v0.3.0 ships 6 checks that detect AI-agent-era manifests on your origin (MCP Server Card, A2A agent-card.json, RFC 9727 API Catalog, agent-permissions.json, OAuth Authorization Server Metadata, WebMCP form annotations). These are informational in v0.3.x — they appear in the report but don't move your score, so the SEO/AEO/GEO baseline (0-100) stays comparable across versions. Future v0.4+ may roll them into scoring once standards stabilize.
See it run
examples/basic-nextjs is a complete, committed Next.js 15 App Router app that consumes every @answerfox/* package, the equivalent of what answerfox init writes. Audited out of the box it scores 79/100; the unshipped checks and the example's minimal content are the ceiling, a content-rich production site scores higher. Clone it, run the audit, read each file.
What's next: Agent Readiness
SEO, AEO, and GEO are about being read by search and answer engines. The shift now underway is being usable by AI agents that act on a user's behalf. Answerfox is adding an Agent Readiness audit: does your site ship the machine-readable manifests agents use to discover and act on it, such as an MCP Server Card, agent-card.json, an API catalog (RFC 9727), agent permissions, OAuth discovery, and WebMCP form annotations?
We scanned Stripe, Vercel, Supabase, and Linear: each is missing 4 to 5 of the 6. Almost nobody has shipped these yet. Like the rest of Answerfox, the plan is to audit your codebase and hand you the manifests as files you commit, not patch you at the edge.
Status: in progress, not yet shipped. If you want your site checked for agent readiness, open an issue with your URL and I'll take a look.
Status
v0.2.x is live on npm under the @answerfox/* scope; CI is green on every package (see the badge above). The audit engine ships 33 of 50 checks. audit and explain are stable and CI-friendly; init and add work interactively in a real terminal.
The remaining 17 checks and the Agent Readiness audit land incrementally. Star and watch the repo to follow releases. The web docs site is in progress; for now, answerfox explain <id> gives the full doc for any check in your terminal.
Architecture
- Monorepo: Turborepo + pnpm workspaces.
- Strict TypeScript:
noUncheckedIndexedAccess,exactOptionalPropertyTypes, noany. Public APIs use branded types so astringcannot pass where anAbsoluteUrlis required. - Eager validation: inputs validate up front with stable error codes (
InvalidUrlError, batchedSchemaValidationError). - Test seams: the CLI's
FsandPrompterare interfaces with in-memory implementations;audit()is a thin fetch wrapper over a purerunChecks(). Tests never hit the network. - Releases: Changesets + Trusted Publishing (OIDC, no tokens).
Documentation
- AUDIT-FRAMEWORK — all 50 checks with rationale
- ROADMAP — phase breakdown
- CONTRIBUTING — set up the workspace and propose changes
Contributing
The fastest way to help: run an audit on a production site you know well, find a result that surprised you (false positive, false negative, or an obvious missing check), and open an issue with the URL. Code contributions welcome, see CONTRIBUTING.md.
License
MIT 2026 Anuj Ojha. The community version stays complete and free. Hosted conveniences (continuous monitoring, history) may come later; the MIT core stays untouched.