Five agent repos gaining momentum. Every one measured.

Ranked by stars gained over the last 7 days, growth rate and recent commit activity.

Rank 1

ChromeDevTools/chrome-devtools-mcp

chrome-devtools-mcp is an MCP server and CLI that lets coding agents (such as Claude, Cursor, Copilot) control and inspect a live Chrome browser. It supports performance trace recording, network request and console analysis, screenshots, and Puppeteer-based automation, and ships with agent skills for tasks like accessibility and LCP debugging.

Developers building coding agents that need to automate, debug, or profile web pages in a real Chrome instance should look at it, noting that browser content is exposed to the client and usage statistics are collected by default unless opted out.

★ 54,313
+★ 397 in 7 days
0.7% growth
Apache-2.0
README ≈ 3,655 tokens to load into context

AI-written summary

Score93.0

Rank 2

oraios/serena

Serena is an MCP toolkit that gives coding agents IDE-style capabilities: symbol-level code retrieval, editing, refactoring and debugging. It runs on language servers (LSP) by default or on a paid JetBrains plugin, and connects to MCP-capable clients such as Claude Code, Codex, Copilot CLI, and desktop or IDE-based assistants.

Developers building or running coding agents on larger codebases who want symbol-aware navigation and cross-file refactors, rather than line-number or text-search edits, should look at it.

★ 30,790
+★ 263 in 7 days
0.9% growth
No licence declared
README ≈ 5,408 tokens to load into context

AI-written summary

Score89.0

Rank 3

activepieces/activepieces

Activepieces is an open-source, self-hostable workflow automation platform positioned as a Zapier replacement, with a no-code builder supporting loops, branches, retries, HTTP, and NPM-based code steps. Integrations are TypeScript npm packages called "pieces," and all pieces are exposed as MCP servers usable from Claude Desktop, Cursor, or Windsurf.

Developers who want to give LLM agents access to many third-party services via MCP, or write and publish their own TypeScript integrations, can use it as a toolkit and a framework.

★ 25,255
+★ 113 in 7 days
0.4% growth
No licence declared

AI-written summary

Score76.0

Rank 4

github/github-mcp-server

GitHub's official MCP server, which connects AI tools to GitHub so agents can read repositories and code, manage issues and pull requests, monitor GitHub Actions runs, and review security findings and Dependabot alerts. It is available as a GitHub-hosted remote server (OAuth or PAT) and as a local version, and it is also published as a Go module (github.com/github/github-mcp-server/v2).

Developers building agents that need to act on GitHub repos, issues, PRs, or CI workflows through an MCP host such as VS Code, Claude Desktop, Cursor, or Windsurf should look at it.

★ 34,017
+★ 210 in 7 days
0.6% growth
MIT

AI-written summary

Score70.0

Rank 5

assafelovic/gpt-researcher

GPT Researcher is an open-source research agent that uses planner and execution agents to break a task into research questions, gather and filter information from web and local sources in parallel, and publish a cited report. It supports multi-agent workflows and recursive deep research, and can be used as a Python package or a Claude Skill.

Developers building agents that need automated, source-cited research over web or local documents can use it as a ready-made planner/executor/publisher pipeline or as a reference architecture.

★ 30,318
+★ 111 in 7 days
0.4% growth
Apache-2.0
README ≈ 7,488 tokens to load into context

AI-written summary

Score52.0