Skip to content

Starred AI / Agent Repositories Over 10K Stars

What it is

This page summarizes the AI and agent-related repositories from your GitHub stars that currently have more than 10,000 GitHub stars. It is meant to answer a practical question: what does each repo actually add to a stack, when should it be used, and which ones are baseline additions versus situational choices.

Star counts below are from a GitHub API snapshot pulled on 2027-01-06 from your starred repositories. "Reputation" is an editorial assessment based on maintainer track record, institutional backing, and ecosystem trust, not a GitHub API field.

What problem it solves

  • Library Overload: Helps navigate the "sea of stars" by filtering for high-momentum, high-reputation projects.
  • Integration Friction: Identifies which tools are "baselines" (always use) vs "situational" (only use for specific tasks).
  • Stack Optimization: Suggests bundles (e.g., Claude-centric, Local-first) to simplify architectural decisions.
  • SOTA Alignment: Integrates with early January 2027 state-of-the-art architectures (Claude 5.6, GPT-5.6, Llama 4, Gemma 4, Qwen 3.6 VL, Gemini 4.0 Ultra series) and the Model Context Protocol (FastMCP 3.1 Task Protocol) to avoid stale recommendations.

Where it fits in the stack

Landscape Intelligence. This is a meta-documentation layer that governs tool selection for all other layers (Infrastructure, Development, Services, and Agents).

Typical use cases

  • Architecting a New Agent: Deciding whether to use a visual builder like Flowise or a code-first framework like OpenCode.
  • Benchmarking Tools: Comparing star counts and reputation to assess the long-term viability of a dependency.
  • Skill Expansion: Identifying high-quality first-party resources (like Anthropic Cookbooks) to improve agent performance.
  • MCP Server Discovery: Programmatically locating baseline repository frameworks supporting the FastMCP 3.1 Task Protocol.

Quick take

Strengths

  • Curated and Prioritized: Focuses only on high-momentum projects (>10K stars).
  • Practical "Default Stance": Provides an immediate "Yes/No/Situational" recommendation for every tool.
  • Ecosystem Awareness: Highlights combinations and bundles that work well together.
  • SOTA Frontier Native: Incorporates early January 2027 multi-modal and structural frameworks natively.

Decision table

Repo Stars Reputation Core value Use when Default stance Best combinations
Significant-Gravitas/AutoGPT 186,050 Early category-defining OSS project; high name recognition, mixed practical fit Popularized autonomous-agent loops and agent-platform thinking You want to study agent history or need a broad autonomous-agent platform Situational Pair with mem0 and browser-use
anomalyco/opencode 131,210 Strong breakout OSS coding-agent project Full coding-agent runtime with modern CLI workflow You want a serious open coding-agent environment Situational primary stack Pair with upstash/context7, anthropics/skills
openai/whisper 106,450 Top-tier lab, widely trusted Speech-to-text layer for voice, meetings, media, and multimodal ingestion Your agent/app needs audio input or transcription Pair with others Pair with n8n, browser-use
anthropics/skills 111,050 Top-tier lab; first-party reference Canonical reusable skill format and examples for Claude-centric workflows You are using Claude Code or Claude-based agents Always consider for Claude stacks Pair with obra/superpowers, upstash/context7
obra/superpowers 101,150 High-signal solo maintainer with major ecosystem adoption Strong engineering process for coding agents: brainstorming, planning, TDD, review You want higher-quality agent output instead of raw speed Always consider for coding agents Pair with anthropics/skills, anthropics/claude-cookbooks
browser-use/browser-use 96,820 Established OSS agent-tool project Makes websites operable by agents when APIs are missing or insufficient The workflow depends on real browser interaction Pair with others Pair with Playwright, Tavily, mem0
getcursor/cursor 86,100 Industry-leading AI-native IDE VS Code fork with deep AI integration and multi-file editing You want the best-in-class AI coding experience in an IDE Baseline for IDE-first workflows Pair with anthropics/skills, superpowers
FlowiseAI/Flowise 56,150 Established OSS vendor in LLM tooling Visual builder for agents, RAG, and workflows You want a visual control plane rather than a pure-code framework Situational primary stack Pair with LocalAI, Supabase, n8n
mem0ai/mem0 53,250 Focused startup with clear category fit Durable memory layer for agents Your agents need persistent user/task/project memory Pair with others Pair with browser-use, deer-flow
upstash/context7 51,450 Strong devtools company reputation Fresh documentation/context retrieval for LLMs and coding agents Agents need current library docs instead of stale model memory Always consider for coding agents Pair with opencode, Claude Code
mendableai/firecrawl 49,150 Breakout web data acquisition for LLMs Turns entire websites into clean Markdown/LLM-ready data Agents need comprehensive web context from JS-heavy sites Baseline for web-heavy agents Pair with browser-use, Tavily
ComposioHQ/awesome-claude-skills 46,450 Strong org and broad ecosystem presence Discovery index for Claude skill packs and workflow resources You want to survey the skills ecosystem quickly Pair with others Pair with anthropics/skills
mudler/LocalAI 45,950 Well-known OSS local-AI project OpenAI-compatible local inference stack You want local/self-hosted inference with broad compatibility Situational primary stack Pair with llmfit, Flowise
msitarzewski/agency-agents 44,750 Strong community-maintained prompt/agent pack A packaged "AI agency" roster of specialist agents You want pre-shaped specialist roles for marketing/content/ops Situational Pair with n8n, Claude Code
anthropics/claude-cookbooks 39,250 Top-tier lab; first-party Concrete implementation examples for building with Claude You want to build with Claude from examples Always consider for Claude stacks Pair with anthropics/skills
google/langextract 37,150 Top-tier lab Structured extraction with grounding and visualization You need reliable extraction from unstructured text Pair with others Pair with NotebookLM, Supabase
karpathy/autoresearch 36,150 Top-tier individual researcher Minimal, research-oriented autonomous iteration loop You want to study compact research-agent loops Situational Pair with deer-flow ideas
wshobson/agents 33,550 Strong solo maintainer with practical Claude Code focus Multi-agent orchestration and subagent workflows for Claude Code You want ready-made multi-agent structure on top of Claude Code Situational primary stack Pair with anthropics/skills, superpowers
bytedance/deer-flow 32,850 Top-tier product organization Researches, codes, and creates through a super-agent harness You want a modern research/coding harness Situational primary stack Pair with Tavily, mem0
gsd-build/get-shit-done 32,450 Strong emerging org Meta-prompting and spec-driven development system for Claude Code You want a lighter-weight alternative to Superpowers-style rigor Situational Pair with anthropics/skills
musistudio/claude-code-router 31,950 Strong community utility Model/provider routing layer for Claude Code You want to keep the Claude Code UX while swapping model backends Pair with others Pair with cc-switch, LocalAI
farion1231/cc-switch 30,450 Strong community desktop utility Desktop control plane for switching across Claude Code, Codex, and Gemini CLI You actively use multiple coding agents Pair with others Pair with claude-code-router
Fosowl/agenticSeek 27,850 Strong solo maintainer, local-first niche Fully local autonomous agent positioned as a local Manus-style system You want local-only autonomous browsing/coding Situational primary stack Pair with LocalAI, llmfit
googleworkspace/cli 22,650 Top-tier platform team Single CLI for Workspace automation across Drive, Gmail, Calendar, etc. Google Workspace is a core operating surface for your business Always consider for Workspace-heavy ops Pair with n8n, Gemini Canvas
stitionai/devika 21,850 Well-known open-source Devin-style project Open-source "agentic software engineer" stack You want a self-hosted autonomous SWE agent product surface Situational primary stack Pair with context7
AlexsJones/llmfit 18,950 Strong individual utility maintainer Hardware/model fit calculator across many providers and models You need to know what can run on your machine Pair with others Pair with LocalAI, Ollama
shanraisshan/claude-code-best-practice 18,450 Community-maintained playbook repo Practical conventions and prompts for using Claude Code more effectively You want field-tested workflow ideas Pair with others Pair with superpowers
plandex-ai/plandex 17,450 Focused OSS vendor Open-source AI coding agent aimed at real-world, large-project work You want a terminal-first coding agent designed for bigger codebases Situational primary stack Pair with context7

Limitations

  • Snapshot-based: Star counts and "reputation" change over time; requires periodic refreshes.
  • Subjective Assessment: "Reputation" is an editorial assessment, not a purely objective metric.
  • High-Bar Filter: May miss smaller, high-quality projects that haven't hit the 10K mark yet.
  • FastMCP 3.1 Migration Overhead: Adapting legacy tools to support explicit tool binding schema constraints takes development time.

What not to overuse

  • Do not default to heavyweight autonomous-agent platforms such as AutoGPT, Devika, AgenticSeek, or DeerFlow unless the task truly needs end-to-end autonomy.
  • Do not use browser automation first when a stable API exists.
  • Do not add memory systems by default; memory is useful only when persistence beats complexity.
  • Do not confuse curated lists and best-practice repos with production architecture. They are accelerators, not substitutes for design.

When to use it

  • When planning a new project or refactoring an existing agent stack.
  • To onboard new developers/agents to the preferred repository patterns of this homelab.
  • During quarterly stack reviews to ensure dependencies are still industry-standard.

What I would actually default to

When not to use it

  • For highly niche tasks where the best tool might be a 100-star specialist repo.
  • When searching for bleeding-edge research that hasn't gained mass adoption yet.

Example company bundles

  • AI operations baseline: Claude skills + Superpowers + Context7 + n8n.
  • Research-heavy company: DeerFlow + Tavily + Browser Use + mem0 + Workspace CLI.
  • Local-first company: LocalAI + llmfit + Ollama + Flowise for internal tools and prototypes.

Getting started

[!NOTE] This is a meta-documentation page for landscape intelligence. To "get started" with these repositories, refer to the individual tool pages or the Sources / references links below. The recommended first step is to audit your current stack against the "Default baseline" recommendations.

CLI examples

[!NOTE] CLI usage varies by repository. Common examples include using npx to run agent utilities or pip to install frameworks. 

# Example: Discovering actions via the Zapier SDK CLI (managed repo)
zapier-sdk list-actions slack

# Example: Installing a coding agent skill
npx @anthropics/skills install

# Example: Starting a FastMCP 3.1 task daemon
mcp-cli run --host 127.0.0.1 --port 8080 --task-protocol mcp-3.1

API examples

[!NOTE] Most of these repositories expose APIs or are used within agentic applications. Refer to canonical pages for specific snippets. 

# Example: Using a memory layer (mem0) in an agent application with Pydantic v2 validation
from pydantic import BaseModel, Field, ConfigDict
from mem0 import Memory

class MemoryItem(BaseModel):
    model_config = ConfigDict(extra="forbid")

    text: str = Field(..., description="Fact or instruction to persist")
    user_id: str = Field(..., description="Target user identifier")

item = MemoryItem(text="User prefers Claude 5.6 for complex architecture tasks", user_id="jules")
m = Memory()
m.add(item.text, user_id=item.user_id)

# Example: Querying the FastMCP 3.1 server capability discovery endpoint
import httpx
response = httpx.get("http://localhost:8080/mcp/3.1/capabilities")
print(response.json())

  • Claude 5.6 — primary reasoning model for these repositories.
  • AI Tool Access Matrix — real-time availability and status of these tools.
  • Model Routing Guide — deciding which model to use with these repositories.
  • Agentic Workflows — patterns for operationalizing these repos.
  • Claude Code — a primary consumer of many of these tools.
  • n8n — the automation engine often used to bridge these libraries.
  • Skills Index — the functional capabilities these repos provide.
  • Architecture Overview — how these tools fit into the global stack.
  • Aider — terminal-based AI coding.
  • Zed — high-performance AI editor.
  • Tabnine — privacy-first AI completions.

Sources / references

Contribution Metadata

  • Last reviewed: 2027-01-07
  • Confidence: high