Article Complete
    AIM

    Claude Code Finally Fixes the Huge Issue With MCPs

    Claude Code Finally Fixes the Huge Issue With MCPs

    Discover how Claude Code's new MCP Tool Search feature enhances performance. Click to learn more about this innovative update!

    Claude Code Finally Fixes the Huge Issue With MCPs
    Supreeth Koundinya·
    5 min

    Thariq Shihipar (Thariq), a member of technical staff at Anthropic, announced that the company is rolling out a new feature to Claude Code called MCP Tool Search. 

    This feature allows Claude Code to dynamically load tools into context when MCP tools would otherwise take up a lot of context, he said in a post on X. 

    “Claude Code detects when your MCP tool descriptions would use more than 10% of context,” he said. “When triggered, tools are loaded via search instead of preloaded.”

    He also stated that this resolves one of their most requested features on GitHub, lazy loading on MCP servers. “Users were documenting setups with 7+ servers consuming 67k+ tokens.”

    Anthriopic has also released the documentation to help developers implement the feature. 

    Responding to the announcement on X, one developer asked Thariq whether Claude is given both regex-based and natural-language search capabilities, and how the model decides when to search for a tool versus assuming none is available. They also asked whether Claude retries if the first search does not surface the right tool.

    To this, Thariq said Anthropic implemented a custom search tool because Claude Code runs entirely client-side. 

    “Basically, it's a regular tool; it takes in a keyword and does a similarity search on it to return a list of tool names,” he wrote. Because it operates like any other tool in the agent loop, Claude can issue multiple searches, refine keywords, or make parallel calls to locate the most relevant MCP tool.

    Another user reacted, “This fixes one of the most painful scaling issues with MCP setups. Was running 5 servers and watching context evaporate before any actual work began.”

    Claude Code, a terminal-centric AI coding assistant, has emerged as one of Anthropic’s most popular products, if not the most successful across the entire AI ecosystem. 

    As of October 2025, Anthropic reported that approximately 1,15,000 developers used Claude Code weekly. In December, the company said it had reached a $1 billion annualised run-rate just six months after making the product publicly available.

    Recently, Anthropic also announced Claude Cowork, a productivity tool designed to extend Claude’s capabilities beyond coding into automating non-technical workflows. 

    Your reaction

    Discussion

    What's your take on this story?

    0/1000

    No comments yet — be the first to share your view.

    Explore our newsletters

    Build your routine with some of our top newsletters or view them all here.

    Wake up informed

    Make sense of the day's AI news and breakthroughs with our morning briefing.

    Belamy
    WeeklyBelamySee the latest

    Industry intelligence

    Receive a roundup of AI adoption stories by industry vertical, curated for professionals.

    3x WeeklySector6See the latest

    By signing up, you agree to our Privacy Policy.

    Membership

    Stay ahead with AI insights.

    Unlimited access to premium articles, exclusive analysis, and the stories shaping India's AI economy.

    Subscribe now

    AIM Americas

    1190 Miraloma Way Suite P, Sunnyvale, CA 94085

    AIM India

    1st Floor, Sakti Statesman, Bellandur, Bengaluru – 560103

    © 2026 AIM. All rights reserved.