Webify cuts web research tokens for coding agents
Claude Code web lookups can become expensive when many agents perform for a simple question. Each opened page can add about 5,000 to 50,000 tokens to the , even though much of that text may not be used. Webify avoids sending whole pages into the agent.
It turns pages into a and retrieves only the parts that match the question. The then receives about 250 to 750 tokens of focused information, with more added if needed. The claimed result is close to accuracy, with the main weakness being less complete coverage on very broad topics.
Webify works with MCP-compatible coding tools, uses search, construction, and small-model synthesis, and is under the MIT License.
Key points
- WebFetch can spend many tokens by putting whole web pages into the .
- Webify retrieves only the page parts that match the question.
- The target input size is about 250 to 750 tokens instead of thousands or tens of thousands.
- It works with MCP-compatible coding tools.
- It is under the MIT License.