Pithelo

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

xerj-org/xerj

XERJ is a local-first search engine and MCP server, shipped as a single Rust binary, that indexes code, docs, logs, JSONL/CSV and PDFs for BM25, vector and hybrid search behind an Elasticsearch-compatible API. It also provides code search, agent memory and a 'reference coding' workflow where an agent clones and indexes related open-source repos to search before writing code.

Developers building coding agents with Claude Code, Cursor or other MCP clients may want to look at it as a way to give agents cited local search and memory instead of having them read files one by one.

  • ★ 18,689
  • +★ 3,061 in 7 days
  • 19.6% growth
  • Apache-2.0
  • README ≈ 17,639 tokens to load into context

AI-written summary

Score90.7
Rank 2

Graphify-Labs/graphify

Graphify is a /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI that maps a codebase and its docs, SQL schemas, configs, PDFs, images, and video into a queryable knowledge graph. Code is parsed locally with tree-sitter AST (deterministic, no LLM), and each edge is tagged EXTRACTED or INFERRED, with no embeddings or vector store.

Developers building coding agents or using AI assistants on large repos can look at it as a graph-traversal alternative to grepping files or vector retrieval.

  • ★ 140,660
  • +★ 5,226 in 7 days
  • 3.9% growth
  • Apache-2.0
  • README ≈ 13,939 tokens to load into context

AI-written summary

Score88.7
Rank 3

vectorize-io/hindsight

Hindsight is an open-source agent memory system that runs as a server (Docker, pip, or Helm) with Python and TypeScript clients, retaining, recalling, and reflecting on stored information so agents learn over time rather than just replaying conversation history. It works with 25+ LLM providers, stores data in PostgreSQL or Oracle, and offers an LLM wrapper, integrations, and an MCP server.

Developers building agents that need long-term memory beyond RAG or knowledge graphs can look at it, especially given its reported LongMemEval results and the live benchmarks it publishes.

  • ★ 57,781
  • +★ 3,957 in 7 days
  • 7.4% growth
  • MIT
  • README ≈ 10,046 tokens to load into context

AI-written summary

Score88.0
Rank 4

paperclipai/paperclip

Paperclip is an open-source Node.js server and React UI for orchestrating teams of AI agents, where you assign goals, set per-agent models and harnesses, and track tasks, budgets, permissions, and history from one dashboard. It works with agents such as OpenClaw, Claude Code, Codex, Cursor, and Gemini CLI, and can be extended through custom processes, HTTP endpoints, or external adapter packages.

Developers running several different coding or task agents who want a single place to coordinate them with org structure, budgets, and governance should look at it.

  • ★ 109,984
  • +★ 2,619 in 7 days
  • 2.4% growth
  • MIT
  • README ≈ 13,523 tokens to load into context

AI-written summary

Score78.0
Rank 5

NousResearch/hermes-agent

The agent that grows with you

  • ★ 259,573
  • +★ 2,460 in 7 days
  • 1.0% growth
  • MIT
  • README ≈ 6,357 tokens to load into context

AI-written summary

Score68.7