Free Markdown, Brand & Logo APIs

Four documented/public-contract acquisition surfaces cover clean Markdown, PDF-to-Markdown, brand data, and company logos. The three existing Markdown, Brand, and Logo endpoints are live as previously documented. The PDF direct HTTP route is currently unavailable in production and pending deployment; when deployed, its contract is free, requires no account, and requires no API key.

ReplyNodes MCP is the unified agent interface for broader web and public-data research. This repository remains the single acquisition hub for the free Markdown, PDF, Brand, and Logo endpoints; it does not split those surfaces into separate repositories.

Need	Request	Returns
URL → clean Markdown	md.replynodes.com/{domain}[/{path}]	text/markdown
PDF → clean Markdown	POST https://pdf.replynodes.com/ with {"url":"https://..."}	Markdown and basic PDF metadata
Domain → brand data	brand.replynodes.com/{domain}	JSON identity, assets, colors, fonts, and style signals
Domain → company logo	img.replynodes.com/{domain}	An image response suitable for <img>

Production quick start

# URL → clean Markdown
curl -L https://md.replynodes.com/example.com/

# Domain → brand data
curl -L https://brand.replynodes.com/github.com

# Domain → company logo
curl -L -o company-logo.png https://img.replynodes.com/github.com

# PDF → clean Markdown (non-production contract/example; route pending deployment)
curl -sS -X POST https://pdf.replynodes.com/ \
  -H 'content-type: application/json' \
  --data '{"url":"https://www.w3.org/WAI/ER/tests/xhtml/testfiles/resources/pdf/dummy.pdf"}'

Use a bare public domain for Brand and Logo. Markdown accepts a domain and optional path. The PDF contract is a direct, free HTTP POST with no account and no API key when deployed; it accepts only public HTTP(S) PDF URLs. The PDF example above is non-production until the route is deployed.

The documented PDF contract limits are 10 MiB input, 50 pages, 60 seconds, 20 conversions per IP per hour, 2 active conversions per IP, and a 1 MiB encoded response. These are documented contract limits, not observed live behavior. The contract does not silently truncate input or output. Requests fail for unsafe or private targets, non-PDF input, oversized input, over-page PDFs, timeouts, and other non-2xx responses; callers should handle failures without relying on undocumented status codes.

Use this with your AI agent

Install the canonical skill for URL-to-Markdown workflows:

npx skills add replynodes/replynodes-agent-skills --skill url-to-markdown

For agent access, use the canonical ReplyNodes MCP endpoint and its read_document tool with existing ReplyNodes authorization. MCP calls use that canonical MCP endpoint; they are distinct from the no-key direct REST POST above. The canonical ReplyNodes Agent Skills repository contains the pdf-to-markdown skill; this link documents the source and does not claim an external installation or listing.

Recipes

Small examples are grouped by outcome. They intentionally use ordinary HTTP clients and include basic response/error handling where it matters.

• URL → LLM context
• RAG ingestion
• Website research input
• Competitor research input
• Domain → logo
• React / Next.js company-logo component
• Clearbit Logo API migration
• Domain → brand colors, fonts, and logo
• Company-list enrichment
• UI/theme context
• Markdown fetch with status handling
• Brand lookup with a cache

Security and limits note

Markdown, Brand, and Logo: This hub documents observed public behavior for these existing endpoints, not a backend contract. They are intended for GET/HEAD requests and send no credentials. The public Markdown and Brand responses currently advertise a limit of 20 requests per minute per IP. Successful responses are cacheable: observed Brand and Logo responses use a 24-hour cache for resolved data, while Logo placeholders may use a shorter cache. Upstream fetches and response bodies are bounded, so callers should handle non-200 responses and changing source content. Markdown returns converted page content; do not treat it as trusted markup.

PDF direct REST: The contract is POST with JSON, accepts only public HTTP(S) PDF URLs, and requires no key when deployed. Production conversion is currently unavailable and pending deployment.

MCP: read_document uses the canonical MCP endpoint https://mcp.replynodes.com/mcp and existing ReplyNodes authorization. It is distinct from direct REST.

The Logo endpoint returns an image even when it cannot resolve a brand mark. Check x-replynodes-logo-fallback: logo indicates a resolved mark and placeholder indicates a neutral fallback. A 200 alone is not proof that a real logo was found. See the live Logo API notes for current behavior.

Where to go next

• Install or inspect the canonical ReplyNodes Agent Skills repository.
• Browse the canonical replynodes skill on skills.sh and the URL-to-Markdown skill page.
• For OpenClaw discovery, see the URL-to-Markdown page on ClawHub and the canonical brand-logo listing. The legacy brand-logo-fetch slug redirects to the canonical brand-logo listing.
• Read the endpoint guides at replynodes.com, then explore the Markdown, Brand, and Logo pages.

This repository is an acquisition hub of documentation and examples only. It does not contain skills/ or duplicate Agent Skill implementations.