Home / Companies / Parallel Web Systems / Blog / Post Details
Content Deep Dive

How to add web search to the Claude Agent SDK (Python and TypeScript)

Blog post from Parallel Web Systems

Post Details
Date Published
Author
Parallel
Word Count
2,073
Company Posts That Month
16
Language
English
Hacker News Points
-
Post removed?
No
Summary

Claude Agent SDK provides the Claude Code agent loop, built-in tools, permissions, and MCP support for Python and TypeScript, including Anthropic’s WebSearch and WebFetch tools, whose search backend cannot be changed and whose page-reading process is intentionally filtered through model-generated prompts. The guide explains how to replace or supplement these built-in tools with Parallel’s remote Search MCP server, which exposes search and multi-page fetch tools through MCP, can operate without an API key at limited rates, and works with providers such as Amazon Bedrock where Anthropic’s server-side WebSearch is unavailable. Python and TypeScript examples show how to configure the MCP server, authorize its tools, disable WebSearch and WebFetch, and isolate SDK settings, while noting that allowing tools does not remove other default Claude Code capabilities such as shell access. The comparison emphasizes differences in result format, fetching capacity, configurability, setup, and pricing, with Anthropic search priced at $10 per 1,000 searches plus token usage and Parallel offering different free and paid options. It recommends evaluating total turns, token costs, reliability, and search quality on a specific workload, using built-in tools for minimal setup and an MCP search service when greater provider flexibility, bulk page retrieval, configurable search settings, or lower search costs are important.

Trends Found in this Post
Trend Post Mentions Total Month Mentions Posts Companies MoM
MCP 27 4,908 480 161 -53%
Use This Data

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.