Pithelo

Five agent repos gaining momentum. Every one measured.

Warming up: ranked by stars per day since creation.

Short candidate pool today: 5 of 5 picks were filled with relaxed rules (5 by star gains since the last snapshot).

Rank 1

wshobson/agents

wshobson/agents is a collection of 94 plugins (202 agents, 185 skills, 105 commands) written in a shared Markdown source and adapted for Claude Code, Codex CLI, Cursor, OpenCode, Antigravity CLI, GitHub Copilot, and Pi. It installs through each tool's plugin marketplace, via `make` targets from a clone, or as individual skills through `gh skill` or `npx skills`.

Developers who want ready-made specialist agents and skills (Python, JavaScript/TypeScript, code review, security scanning, and more) usable across several AI coding tools without rewriting them per harness should look at it.

  • ★ 41,183
  • MIT
  • README ≈ 3,492 tokens to load into context

AI-written summary

Score48.1
Rank 2

mem0ai/mem0

Mem0 is a memory layer for AI agents and apps that stores and retrieves User, Session, and Agent state so context persists across interactions. It is available as a Python/npm library, a self-hosted server, or a managed cloud platform, with retrieval that combines semantic, BM25 keyword, and entity matching plus time-aware ranking.

Developers building agents, assistants, or support bots that need to recall user preferences and history across sessions should look at it, especially since the README publishes benchmark results (LoCoMo, LongMemEval, BEAM) and an open-sourced evaluation framework to reproduce them.

  • ★ 67,487
  • Apache-2.0
  • README ≈ 5,269 tokens to load into context

AI-written summary

Score43.7
Rank 3

bytedance/deer-flow

DeerFlow 2.0 is an open-source "super agent harness" from ByteDance that orchestrates sub-agents, memory, sandboxes, tools, and extensible skills to handle research, coding, and content-creation tasks lasting minutes to hours. It also includes a message gateway/IM channels, MCP server support, and tracing integrations, and is a ground-up rewrite that shares no code with v1.

Developers building long-running, multi-step agents who want a ready-made harness (sandboxing, memory, sub-agents, skills) rather than assembling those pieces themselves should look at it.

  • ★ 84,168
  • MIT

AI-written summary

Score43.2
Rank 4

infiniflow/ragflow

RAGFlow is an open-source Retrieval-Augmented Generation engine that combines RAG with agent capabilities to supply context to LLMs. It provides a RAG workflow with a context engine, pre-built agent templates, and data ingestion from sources such as sitemaps and Google BigQuery.

Developers building production RAG or agentic RAG systems over complex or unstructured documents, with either cloud or local deployment, may want to evaluate it.

  • ★ 92,502
  • Apache-2.0
  • README ≈ 7,054 tokens to load into context

AI-written summary

Score41.0
Rank 5

langchain-ai/langchain

LangChain is a framework for building agents and LLM-powered applications by chaining together interoperable components and third-party integrations. It provides a standard interface for models, embeddings, vector stores, and tools, and connects to related products such as LangGraph, Deep Agents, and LangSmith.

Developers building LLM-based agents who want swappable model providers and a shared integration layer, with optional paths to LangGraph for orchestration and LangSmith for evals and debugging, should look at it.

  • ★ 148,101
  • MIT
  • README ≈ 2,133 tokens to load into context

AI-written summary

Score40.8