GitHub: AI Agents Dominate Trending Repositories in July 2026

GitHub: AI Agents Dominate Trending Repositories in July 2026

If you've spent time on GitHub Trending this month, you've likely noticed a pattern: it's no longer research papers transforming into repositories, but agents. Coding agents, penetration testing agents, trading agents, and the infrastructure that connects them all.

We tracked the growth of stars, momentum, and real impact to identify the ten repositories that mattered most this month. Rather than ranking projects solely by the number of stars, we considered their influence on the AI ecosystem and how quickly they are gaining traction. In this article, we will detail each repository, what it does, why it’s trending, and why it deserves a spot on your watchlist.

1. usestrix/strix (~42K stars)

Strix is an open-source AI penetration testing tool that behaves like a real security researcher rather than a static scanner. It dynamically tests applications, validates vulnerabilities with proof-of-concept exploits, and includes features such as an HTTP proxy, browser exploitation, a Python sandbox, and CI/CD integration. Its rapid growth, adding about 7,000 stars per week, suggests it is genuinely being adopted by security teams rather than just attracting stars as a passing trend.

• Security teams looking for continuous, AI-driven penetration testing in CI/CD
• Developers needing proof-of-concept validation rather than noisy static analysis alerts
• Engineers exploring how agentic AI applies to offensive security

2. xai-org/grok-build (~9.3K stars)

Grok Build is the command-line interface and terminal interface of xAI, powering the same agent loop behind the Grok coding stack. Released under the Apache 2.0 license, it offers complete transparency on context management, tool execution, plugins, skills, and MCP integration. Although xAI does not accept external contributions, developers can study, compile, and run the agent locally, making it one of the most significant open-source AI outputs of the month.

• Engineers wanting to study a production-quality coding agent line by line
• Teams building their own agent tools and looking for a proven reference architecture
• Anyone following how leading labs approach local agent infrastructure

3. HKUDS/Vibe-Trading (~24K stars)

Developed by the Data Science Lab at the University of Hong Kong, Vibe-Trading converts natural language prompts into backtests, alpha benchmarks, and optional live trades through supported brokers. It includes 452 pre-built alpha factors, point-in-time data management to prevent lookahead bias, and rigorous validation techniques that set it apart from typical AI trading bots.

• Developers curious about quality research backtesting and alpha framework
• Traders exploring natural language-driven strategy research before taking action
• Anyone studying how academic labs approach agentic financial tools

4. DeusData/codebase-memory-mcp (~32K stars)

codebase-memory-mcp is an MCP (Model Context Protocol) server that helps AI coding agents understand large codebases without repeatedly scanning files. It builds a persistent knowledge graph of functions, classes, call chains, and routes using tree-sitter across 158 languages, reducing token usage for structural queries by up to 99%. Distributed as a single static C binary with no dependencies, it runs entirely locally and can index even massive repositories, including the Linux kernel, in just minutes.

• Anyone whose AI coding agent consumes excessive tokens while exploring large codebases
• Teams standardizing on MCP-based tools for Claude Code, Cursor, or similar agents
• Engineers wanting structural code intelligence without running an LLM for every query

5. langchain-ai/openwiki (~11.8K stars)

OpenWiki is a command-line interface from the LangChain team that automatically generates and maintains AI-friendly documentation for your codebase. Although it has fewer stars than some projects on this list, LangChain's influence in the GenAI ecosystem makes it a notable output. OpenWiki helps keep projects understandable for AI agents, making codebases easier to navigate, maintain, and work on over time.

• Teams wanting documentation that an AI agent can reliably consume and act upon
• Engineers standardizing on the broader LangChain agent tools ecosystem
• Anyone maintaining a large codebase where documentation regularly becomes outdated

6. MadsLorentzen/ai-job-search (~23K stars)

Built on Claude Code, this framework automates the application process by evaluating job listings, tailoring resumes, generating cover letters, and preparing candidates for interviews. Although it is a project by a single developer, it has gained popularity by solving a common real-world problem. More than anything, it reflects the broader trend this month: AI agents are increasingly designed to handle practical, everyday workflows rather than merely showcasing new models.

• Job seekers wanting to automate the repetitive parts of applications
• Developers curious about how Claude Code can be forked into a personal-use agent
• Anyone looking for a practical, everyday example of agentic AI in action

7. iOfficeAI/OfficeCLI (~18K stars)

OfficeCLI is a free open-source office suite designed for AI agents to read, edit, and automate Word, Excel, and PowerPoint files, delivered as a single binary with no Office installation required. It follows the same trend as the MCP server tools mentioned elsewhere on this list: making everyday file formats natively readable and editable by AI agents without requiring a human GUI interface. It may not be flashy, but it’s the kind of infrastructure repository that quietly ends up integrated into many automated workflows.

• Teams automating document generation and editing via AI agents
• Developers needing support for Office files without installing Office itself
• Anyone building MCP-based agent workflows around everyday business documents

8. diegosouzapw/OmniRoute (~17.9K stars)

OmniRoute is a free AI gateway that provides a single access point to route queries through over 231 providers, including more than 50 free ones, allowing you to connect tools like Claude Code, Codex, Cursor, and Copilot to a wide range of language models. It integrates token compression, intelligent auto-return, and multimodal API support. It’s a genuinely practical infrastructure tool, though it falls more into the utility category than groundbreaking: the kind of repository you star because it saves you real setup time, not because it changes your thinking about AI.

• Developers wanting a single access point instead of juggling multiple provider API keys
• Teams looking to reduce token costs with compression and intelligent return
• Anyone connecting multiple coding agents to a shared pool of free and paid models

9. JustVugg/colibri (~14.7K stars)

Colibri is a pure C inference engine with zero dependencies that allows you to run GLM-5.2, a 744 billion parameter model, on a consumer machine with about 25 GB of RAM, streaming experts from disk as needed. It’s a true engineering feat concentrated in a small space. Its audience is narrower than most on this list, primarily local LLM enthusiasts and those keen on running state-of-the-art models without cloud infrastructure, but for this audience, it’s a big deal.

• Local LLM enthusiasts wanting state-of-the-art models on consumer hardware
• Engineers curious about expert mixture inference streamed from disk
• Anyone prioritizing privacy and cost control over cloud-based inference

10. Nutlope/hallmark (~10K stars)

Hallmark is a design skill for Claude Code, Cursor, and Codex that counters the generic UI output and over-distribution that most large language models adopt by default. It executes fifty-seven "slop test gates" plus a pre-release...