bolt.new — F/100 agent-readiness

01Fail
Markdown content negotiation
+8 pts if fixed
Serve text/markdown when the client sends Accept: text/markdown, or expose /path/index.md.
See full details
02Fail
Explicit AI bot rules
+5 pts if fixed
Add explicit allow/deny rules for GPTBot, ClaudeBot, PerplexityBot, Google-Extended, etc.
See full details
03Fail
API Catalog (RFC 9727)
+4 pts if fixed
Publish /.well-known/api-catalog linking to your OpenAPI descriptions.
See full details

Hand-off prompt for your coding agent6 issues

# Task: raise AIScan agent-readiness score

Site: https://bolt.new/
Platform: bolt (99% confidence)
AIScan score: 31/100 — Level 2 (Readable)
Rubric: v2026-06-08.1

Implement the fixes below in this repository. Make atomic commits per check id. After all edits, run the build to confirm nothing breaks.

## Failing checks
- **[C1] Markdown content negotiation** — FAIL
  - Evidence: Accept: text/markdown → 200 (text/html)
  - Fix: Serve text/markdown when the client sends Accept: text/markdown, or expose /path/index.md.
- **[C2] /llms.txt** — PARTIAL
  - Evidence: 200 OK, 62119 bytes
  - Fix: Publish /llms.txt per llmstxt.org with an H1, summary, and ## sections of markdown links.
- **[B2] Explicit AI bot rules** — FAIL
  - Evidence: 0 known AI agent(s) addressed in robots.txt
  - Fix: Add explicit allow/deny rules for GPTBot, ClaudeBot, PerplexityBot, Google-Extended, etc.
- **[P1] API Catalog (RFC 9727)** — FAIL
  - Evidence: 404 at /.well-known/api-catalog
  - Fix: Publish /.well-known/api-catalog linking to your OpenAPI descriptions.
- **[P4] OAuth discovery metadata** — FAIL
  - Evidence: auth-server: 404, protected-resource: 404
  - Fix: Publish RFC 8414 / RFC 9728 metadata at the /.well-known/oauth-* endpoints.
- **[D3] Link header for discovery** — FAIL
  - Evidence: no Link response header
  - Fix: Add a Link header with rel="api-catalog" or rel="describedby".

## Notes
- Treat `public/` (or framework equivalent) as the source for `robots.txt`, `llms.txt`, and `.well-known/` files.
- For Markdown content negotiation, respond to `Accept: text/markdown` and `?format=md` on canonical content URLs.
- For MCP / Agent Skills, drop card files under `.well-known/` even if the server isn't live yet — discovery matters first.

Discoverability

Can agents find your pages? robots, sitemaps, llms.txt.

Your site is missing this right now, so AI agents can't use it. The fix below is what to change.
ID · D3
What we found
no Link response header
How to fix it
Add a Link header with rel="api-catalog" or rel="describedby".
1.⚡ Fix automatically with AIScan
The fastest way to apply this fix is using AIScan's own agent tooling — it gives any AI coding assistant machine-readable instructions for this exact check. **Option A — Claude Code skill** 1. Download `CLAUDE.md` from https://aiscan.site/CLAUDE.md and drop it into your project root. 2. Ask Claude Code: *"Fix the failing AIScan check for <your-url>"* — it will read the skill, call our API, and apply the fix. **Option B — MCP (any MCP-compatible agent: Claude Desktop, Cursor, etc.)** - Endpoint: `https://aiscan.site/api/mcp` - Tools: `scan_website`, `get_fixes`, `get_grade` - Call `get_fixes` with your URL for machine-readable fix instructions. **Option C — REST API (any tool / script)** ``` GET https://aiscan.site/api/public/scan?url=YOUR_URL ``` Returns the full report including failing checks and fix details. Then re-scan at https://aiscan.site to verify.
2.Add a Link response header
Set this header on your homepage and/or root document. Most CDNs (Cloudflare, Vercel, Netlify) support response-header rules.
3.Example
Link: </.well-known/api-catalog>; rel="api-catalog"
4.Verify
After deploying, re-scan at https://aiscan.site. You can also test with `curl -I` or `curl` directly.
Fix in ChatGPT Fix in Claude

Content

Can agents read your content cleanly? Markdown, structured data.

111

Your site is missing this right now, so AI agents can't use it. The fix below is what to change.
ID · C1
What we found
Accept: text/markdown → 200 (text/html)
How to fix it
Serve text/markdown when the client sends Accept: text/markdown, or expose /path/index.md.
1.⚡ Fix automatically with AIScan
The fastest way to apply this fix is using AIScan's own agent tooling — it gives any AI coding assistant machine-readable instructions for this exact check. **Option A — Claude Code skill** 1. Download `CLAUDE.md` from https://aiscan.site/CLAUDE.md and drop it into your project root. 2. Ask Claude Code: *"Fix the failing AIScan check for <your-url>"* — it will read the skill, call our API, and apply the fix. **Option B — MCP (any MCP-compatible agent: Claude Desktop, Cursor, etc.)** - Endpoint: `https://aiscan.site/api/mcp` - Tools: `scan_website`, `get_fixes`, `get_grade` - Call `get_fixes` with your URL for machine-readable fix instructions. **Option C — REST API (any tool / script)** ``` GET https://aiscan.site/api/public/scan?url=YOUR_URL ``` Returns the full report including failing checks and fix details. Then re-scan at https://aiscan.site to verify.
2.Serve text/markdown on Accept negotiation
When a client sends `Accept: text/markdown`, return a markdown rendering of the same content. The simplest path is to publish parallel `.md` files (e.g. `/about` → `/about.md`) and rewrite based on the Accept header.
3.Test it
curl -H "Accept: text/markdown" https://yoursite.com/
4.Verify
After deploying, re-scan at https://aiscan.site. You can also test with `curl -I` or `curl` directly.
Fix in ChatGPT Fix in Claude

Bot Access

Are AI crawlers explicitly allowed or blocked?

Your site is missing this right now, so AI agents can't use it. The fix below is what to change.
ID · B2
What we found
0 known AI agent(s) addressed in robots.txt
How to fix it
Add explicit allow/deny rules for GPTBot, ClaudeBot, PerplexityBot, Google-Extended, etc.
1.⚡ Fix automatically with AIScan
The fastest way to apply this fix is using AIScan's own agent tooling — it gives any AI coding assistant machine-readable instructions for this exact check. **Option A — Claude Code skill** 1. Download `CLAUDE.md` from https://aiscan.site/CLAUDE.md and drop it into your project root. 2. Ask Claude Code: *"Fix the failing AIScan check for <your-url>"* — it will read the skill, call our API, and apply the fix. **Option B — MCP (any MCP-compatible agent: Claude Desktop, Cursor, etc.)** - Endpoint: `https://aiscan.site/api/mcp` - Tools: `scan_website`, `get_fixes`, `get_grade` - Call `get_fixes` with your URL for machine-readable fix instructions. **Option C — REST API (any tool / script)** ``` GET https://aiscan.site/api/public/scan?url=YOUR_URL ``` Returns the full report including failing checks and fix details. Then re-scan at https://aiscan.site to verify.
2.Add explicit User-agent blocks
Don't rely on `User-agent: *` alone. Address each major AI crawler so your policy is unambiguous.
3.Recommended block
User-agent: GPTBot Allow: / User-agent: ClaudeBot Allow: / User-agent: PerplexityBot Allow: / User-agent: Google-Extended Allow: / User-agent: Applebot-Extended Allow: /
4.Verify
After deploying, re-scan at https://aiscan.site. You can also test with `curl -I` or `curl` directly.
Fix in ChatGPT Fix in Claude

Capabilities

Can agents do things? APIs, auth, MCP, Agent Skills.

Your site is missing this right now, so AI agents can't use it. The fix below is what to change.
ID · P1
What we found
404 at /.well-known/api-catalog
How to fix it
Publish /.well-known/api-catalog linking to your OpenAPI descriptions.
1.⚡ Fix automatically with AIScan
The fastest way to apply this fix is using AIScan's own agent tooling — it gives any AI coding assistant machine-readable instructions for this exact check. **Option A — Claude Code skill** 1. Download `CLAUDE.md` from https://aiscan.site/CLAUDE.md and drop it into your project root. 2. Ask Claude Code: *"Fix the failing AIScan check for <your-url>"* — it will read the skill, call our API, and apply the fix. **Option B — MCP (any MCP-compatible agent: Claude Desktop, Cursor, etc.)** - Endpoint: `https://aiscan.site/api/mcp` - Tools: `scan_website`, `get_fixes`, `get_grade` - Call `get_fixes` with your URL for machine-readable fix instructions. **Option C — REST API (any tool / script)** ``` GET https://aiscan.site/api/public/scan?url=YOUR_URL ``` Returns the full report including failing checks and fix details. Then re-scan at https://aiscan.site to verify.
2.Publish an API catalog (RFC 9727)
# Add to your homepage HTTP response headers Link: </.well-known/api-catalog>; rel="api-catalog" # And serve a JSON document at /.well-known/api-catalog: { "links": [ { "href": "/openapi.json", "rel": "service-desc", "type": "application/openapi+json" } ] }
3.Verify
After deploying, re-scan at https://aiscan.site. You can also test with `curl -I` or `curl` directly.
Fix in ChatGPT Fix in Claude
Your site is missing this right now, so AI agents can't use it. The fix below is what to change.
ID · P4
What we found
auth-server: 404, protected-resource: 404
How to fix it
Publish RFC 8414 / RFC 9728 metadata at the /.well-known/oauth-* endpoints.
1.⚡ Fix automatically with AIScan
The fastest way to apply this fix is using AIScan's own agent tooling — it gives any AI coding assistant machine-readable instructions for this exact check. **Option A — Claude Code skill** 1. Download `CLAUDE.md` from https://aiscan.site/CLAUDE.md and drop it into your project root. 2. Ask Claude Code: *"Fix the failing AIScan check for <your-url>"* — it will read the skill, call our API, and apply the fix. **Option B — MCP (any MCP-compatible agent: Claude Desktop, Cursor, etc.)** - Endpoint: `https://aiscan.site/api/mcp` - Tools: `scan_website`, `get_fixes`, `get_grade` - Call `get_fixes` with your URL for machine-readable fix instructions. **Option C — REST API (any tool / script)** ``` GET https://aiscan.site/api/public/scan?url=YOUR_URL ``` Returns the full report including failing checks and fix details. Then re-scan at https://aiscan.site to verify.
2.Publish OAuth discovery metadata
If your API supports OAuth, serve RFC 8414 metadata at `/.well-known/oauth-authorization-server` and/or RFC 9728 at `/.well-known/oauth-protected-resource`.
3.Verify
After deploying, re-scan at https://aiscan.site. You can also test with `curl -I` or `curl` directly.
Fix in ChatGPT Fix in Claude

Showcase your grade

Embed your AIScan badge

Drop this badge into your site footer, GitHub README, or docs. It links back to a fresh scan of bolt.new on AIScan.site.

HTML

<!-- AIScan.site agent-readiness badge -->
<a href="https://aiscan.site/?u=https%3A%2F%2Fbolt.new%2F" target="_blank" rel="noopener" title="Agent-Readiness Grade F (31/100) — AIScan.site">
  <img src="https://aiscan.site/api/public/badge.svg?score=31&grade=F&host=bolt.new" alt="AIScan.site Grade F — 31/100" width="240" height="64" loading="lazy" />
</a>

Markdown

[![AIScan.site Grade F — 31/100](https://aiscan.site/api/public/badge.svg?score=31&grade=F&host=bolt.new)](https://aiscan.site/?u=https%3A%2F%2Fbolt.new%2F)

Track this over time

Coming soon: save sites, schedule re-scans, get email when something material changes. Standards land weekly — your tracker shouldn't be one-shot.

Free · No signup · ~20 seconds

Scan your own site

Get a plain-English agent-readiness report you can hand straight to ChatGPT or Claude.

Scan your site

Or browse the community feed to see what others have scanned.

Bolt

Quick Wins

Markdown content negotiation

Explicit AI bot rules

API Catalog (RFC 9727)

Discoverability

robots.txt present & sane

XML sitemap

Link header for discovery

Content

Markdown content negotiation

/llms.txt

Structured HTML (title, meta, JSON-LD, single H1)

Bot Access

Content Signals in robots.txt

Explicit AI bot rules

Web Bot Auth key directory

Capabilities

API Catalog (RFC 9727)

MCP Server Card

Agent Skills index

OAuth discovery metadata

Embed your AIScan badge

Track this over time

Scan your own site

Bolt

Quick Wins

Markdown content negotiation

Explicit AI bot rules

API Catalog (RFC 9727)

robots.txt present & sane

XML sitemap

Link header for discovery

Markdown content negotiation

/llms.txt

Structured HTML (title, meta, JSON-LD, single H1)

Content Signals in robots.txt

Explicit AI bot rules

Web Bot Auth key directory

API Catalog (RFC 9727)

MCP Server Card

Agent Skills index

OAuth discovery metadata

Share as a social image

Embed your AIScan badge

Track this over time

Scan your own site