Use case
AI agents with web access
Let an agent read the web as Markdown instead of raw HTML.
The problem.
Raw pages burn tokens. The agent loses track of what mattered. It needs the content as text. Links must stay intact.
How Webfold helps.
Step 1
Add the MCP server
One config block in Claude Desktop, Cursor or any MCP client.
webfold-mcpStep 2
Let it search, then read
The agent searches, picks results, and fetches them as Markdown.
webfold search "topic" --count 10Step 3
Crawl when it needs more
For a docs site or a knowledge base, crawl it and search the result.
webfold crawl https://docs.example.com --max 100Products used.
Worth knowing.
Note
- Treat fetched text as untrusted. It is not a defence against prompt injection.
Try it on your own data.
Run the command-line tool today. Hosted API access is coming.