Empowering AI Agents to browse the web with clarity and intelligence.
The web was built for humans, not machines. While humans can effortlessly parse HTML, navigate complex layouts, and understand visual hierarchies, AI Agents struggle with these tasks. MarkWay bridges this gap by providing a protocol that speaks the language AI understands best: structured, semantic content.
Traditional web browsing requires AI Agents to:
- Parse complex HTML structures
- Execute JavaScript to render content
- Deal with inconsistent class names and DOM structures
- Handle anti-scraping measures and rate limits
- Extract meaningful data from visual layouts
This approach is fragile, inefficient, and computationally expensive.
MarkWay introduces a Markdown-first protocol that transforms the web experience for AI Agents:
Web → Markdown → Agent
By serving content directly in Markdown format, MarkWay enables:
- Semantic clarity: Content structure is explicit and unambiguous
- Efficient parsing: No HTML parsing overhead or DOM manipulation
- Natural comprehension: AI reads like humans read documentation
- Protocol discovery: Built-in mechanism for Agents to understand site structure
MarkWay operates in two modes:
Content is organized as Markdown files with predictable directory structures:
- Every directory contains an
index.mdwith a structured table of contents - All links end with
.mdfor consistency - Supports relative, absolute, and external paths
Interactive endpoints use HTTP methods semantically:
GETreturns Markdown documentation (no URL parameters)POSTperforms data exchange via request body- Response format negotiation via
Acceptheaders
For site owners:
- Serve your content in Markdown format
- Include
index.mdin each directory with a table listing files and their purposes - Root request must link to protocol documentation at
/protocol
For AI Agent developers:
- Detect MarkWay sites via protocol discovery at
/protocolor/protocol.md - Parse
index.mdtables to navigate site structure - Use
GETfor documentation,POSTfor data exchange
The complete protocol specification is available in:
- API Documentation: Human-readable docs that Agents can parse flawlessly
- Knowledge Bases: Structured information accessible to both humans and AI
- Data Services: Query endpoints with self-documenting interfaces
- AI-Native Applications: Build web services designed for Agent consumption
We welcome contributions! Please submit issues and pull requests to help improve the MarkWay protocol.
MIT License
MarkWay: The web, reimagined for the age of AI.