AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

Semble is a code search tool designed for agents that achieves 98% fewer tokens than grep+read, enabling faster and more efficient code retrieval. It indexes repos rapidly and runs entirely on CPU, with no external dependencies. This development could significantly improve code search workflows for AI agents.

Semble, a new code search library tailored for AI agents, claims to cut token usage by approximately 98% compared to traditional grep+read methods, while maintaining high accuracy and speed. It runs entirely on CPU, requires no external services, and can be integrated with popular agents like Claude Code, Codex, and OpenCode, offering instant code retrieval.

Developed for use with agents, Semble indexes a full codebase in about 250 milliseconds and answers queries in roughly 1.5 milliseconds. Benchmarks indicate it achieves a normalized discounted cumulative gain (NDCG@10) of 0.854, comparable to specialized transformer models but at a fraction of their size and cost. The tool can be deployed as an MCP server or used directly via command line, supporting local repositories or remote git URLs.

Semble’s core advantage is its token efficiency, returning only the relevant code snippets with significantly fewer tokens—around 98% less—than traditional grep-based methods. It operates solely on CPU, eliminating the need for GPUs or API keys, which simplifies setup and reduces costs. It also offers features like automatic re-indexing on file changes and caching for session persistence.

Why It Matters

This development matters because it offers a more efficient, faster, and cost-effective way for AI agents to perform code searches, which can enhance automation, debugging, and development workflows. By reducing token consumption and latency, Semble can improve the responsiveness and scalability of code retrieval systems, especially in large codebases.

FOXWELL NT301 OBD2 Scanner Live Data Professional Mechanic OBDII Diagnostic Code Reader Tool for Check Engine Light

FOXWELL NT301 OBD2 Scanner Live Data Professional Mechanic OBDII Diagnostic Code Reader Tool for Check Engine Light

  • Read Fault Codes: Requires ignition on and check engine light on
  • Vehicle CEL Diagnosis: Read DTCs, reset MIL, retrieve VIN
  • Live Data Graphing: Monitor sensors and trends in real time

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Background

Traditional code search methods like grep read entire files and can be token-heavy, limiting efficiency in AI-driven workflows. Existing transformer-based models for code retrieval are accurate but costly and slow, often requiring specialized hardware. Semble’s approach, which emphasizes speed, token efficiency, and local CPU operation, addresses these limitations. Its announcement on Hacker News highlights ongoing efforts to optimize developer tools for AI integration, with similar tools gaining interest in the developer community.

“Semble returns only the relevant chunks, using ~98% fewer tokens than grep+read, with comparable retrieval quality.”

— Semble team

“This could be a game-changer for AI code agents, making code search faster and cheaper.”

— Hacker News user

The No-Code AI Business Toolkit: Tools, templates, workflows and prompts for building smarter business systems without code. (The Practical AI & SEO Business Library)

The No-Code AI Business Toolkit: Tools, templates, workflows and prompts for building smarter business systems without code. (The Practical AI & SEO Business Library)

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

What Remains Unclear

Details about long-term scalability, integration with all agent types, and real-world performance in large, complex codebases remain to be fully validated. It is also unclear how Semble performs across different programming languages and in diverse development environments.

DeskFX Free Audio Effects & Audio Enhancer Software [PC Download]

DeskFX Free Audio Effects & Audio Enhancer Software [PC Download]

  • Audio Transformation: Enhance sound through speakers and headphones
  • Sound Quality Improvement: Adjust audio with various effects
  • Audio Control: Manage sound output from hardware

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

What’s Next

Next steps include broader adoption by AI developer communities, further benchmarking in varied environments, and potential integration into more agent frameworks. Updates may include feature enhancements, support for additional languages, and performance improvements based on user feedback.

Code by Note, Bk 1: Find the Patterns by Reading the Notes, Coloring Book (Color by Note, Bk 1)

Code by Note, Bk 1: Find the Patterns by Reading the Notes, Coloring Book (Color by Note, Bk 1)

  • Condition: Used Book in Good Condition

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Key Questions

How does Semble compare to traditional grep in terms of speed?

Semble indexes repositories in about 250 ms and answers queries in roughly 1.5 ms, making it significantly faster than grep+read, especially for large codebases.

Can Semble be used with existing AI code agents?

Yes, Semble supports integration with agents like Claude Code, Codex, and OpenCode via MCP or command line, enabling instant code search capabilities.

Is Semble suitable for large, multi-language repositories?

While designed for efficiency, performance in large or multi-language repos depends on indexing and setup. Benchmarks suggest good speed and accuracy, but real-world results may vary.

What are the setup requirements for using Semble?

Semble runs on CPU, requires Python and uv, and can be integrated via MCP or CLI. No GPU, API keys, or external services are necessary.

You May Also Like

How GLM-5.3’s Frontier Coding Transformed AI’s Cyber Capabilities

Z.ai’s GLM-5.3 enhances coding capabilities by 50%, but its emergent cybersecurity abilities raise safety and governance concerns.

Candor as a Moat: A Critical Reading of Dario Amodei and Anthropic

Examining how Dario Amodei’s transparency and safety proposals may serve as strategic barriers for Anthropic amid recent regulatory actions.

Understanding Anthropic’s $965B Series H: The Compute Revolution

Anthropic’s latest funding round emphasizes massive investments in compute infrastructure, chips, and power to scale AI models like Claude, marking a shift in AI development priorities.

Did AI Really Intent To Attack? The Surprising Backstory Of Its First Cyberattack

Exploring the first documented autonomous AI cyberattack, its true motive, and implications for AI safety and security.