Let your AI agent read progressively through large pages.

Different AI models have different context windows. A long documentation page that fits one model’s input budget may exceed another’s, causing a context-limit error before the agent can use the content.

Web Data Source (WDS) gives agents a way to read progressively through its MCP tools:

• WdsFetchChunked starts or resumes downloading a page and returns its task ID.
• WdsGetChunk reads a portion of that downloaded content, with a chosen starting position and maximum character count.
• The response reports the full content length, helping the agent plan subsequent reads.

For an agent reviewing a long technical document, the workflow becomes:

Read a portion → Process it → Retain useful findings → Continue.

By choosing suitable portion sizes and summarizing findings between reads, an agent can work through the page while managing its available context. This helps reduce the risk of oversized inputs across models with different limits.

WDS controls each read in characters; the agent manages the model’s token budget and accumulated context.

Large pages. Manageable reads. More flexibility across AI models.

Explore the corresponding WDS MCP tools