The Inference Report

March 15, 2026

The GitHub ecosystem is consolidating around two distinct problems: making agents useful, and making agents work at scale. The agent trend itself isn't new, but the infrastructure supporting it has matured. Repos like msitarzewski/agency-agents (44775 stars) and obra/superpowers (84068 stars) aren't just frameworks anymore; they're methodology. They ship with personality, processes, and deliverables built in, treating multi-agent systems as a solved pattern rather than a research question. Meanwhile, specialized tools like volcengine/OpenViking handle context management for agents through a file system paradigm, and dimensionalOS/dimos extends agents into physical space by binding natural language to hardware actuators and sensors. The common thread: agents need scaffolding. They need memory, they need skills, they need the ability to act on the world. The repos gaining traction are the ones that provide that scaffolding instead of just the agent itself.

Model capability and tooling are decoupling. anthropics/claude-plugins-official marks a shift toward plugin ecosystems as the distribution model for agent capabilities, while unsloth (53998 stars) and camel-ai/camel (16350 stars) show that fine-tuning and multi-agent coordination are now treated as engineering problems, not research problems. The discovery layer reveals where developers are investing energy: autonomous research loops (wanshuiyin/Auto-claude-code-research-in-sleep), dataset quality (argilla-io/argilla), and benchmarking (openml/automlbenchmark). lightpanda-io/browser (17501 stars) solves a concrete problem for AI automation by building a headless browser designed for agent interaction rather than retrofitting Puppeteer. p-e-w/heretic's traction (14120 stars) signals that content moderation and output control remain active concerns. The pattern is clear: developers are moving past "can we build an agent" to "how do we build agents that work reliably in production, integrate with existing systems, and handle real constraints."

Jack Ridley

Trending
volcengine/OpenViking
11197

OpenViking is an open-source context database designed specifically for AI Agents(such as openclaw). OpenViking unifies the management of context (memory, resources, and skills) that Agents need through a file system paradigm, enabling hierarchical context delivery and self-evolving.

anthropics/claude-plugins-official
11518

Official, Anthropic-managed directory of high quality Claude Code Plugins.

dimensionalOS/dimos
981

Dimensional is the agentic operating system for physical space. Vibecode humanoids, quadrupeds, drones, and other hardware platforms in natural language and build multi-agent systems that work seamlessly with physical input (cameras, lidar, actuators).

p-e-w/heretic
14120

Fully automatic censorship removal for language models

langflow-ai/openrag
2855

OpenRAG is a comprehensive, single package Retrieval-Augmented Generation platform built on Langflow, Docling, and Opensearch.

lightpanda-io/browser
17501

Lightpanda: the headless browser designed for AI and automation

msitarzewski/agency-agents
44775

A complete AI agency at your fingertips** - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.

fishaudio/fish-speech
27360

SOTA Open Source TTS

InsForge/InsForge
4329

Give agents everything they need to ship fullstack apps. The backend built for agentic development.

obra/superpowers
84068

An agentic skills framework & software development methodology that works.

Daily discovery
Firmament-Autopilot/FMT-FirmwareRobotics
695

Firmament Autopilot Embedded System

wanshuiyin/Auto-claude-code-research-in-sleepMCP
1068

ARIS ⚔️ (Auto-Research-In-Sleep) — Claude Code skills for autonomous ML research: cross-model review loops, idea discovery, and experiment automation via Codex MCP

OliBomby/MapperatorinatorGenerative AI
431

An AI framework for generating and modding osu! beatmaps for all gamemodes from spectrogram inputs.

unslothai/unslothFine-tuning
53998

Fine-tuning & Reinforcement Learning for LLMs. 🦥 Train OpenAI gpt-oss, DeepSeek, Qwen, Llama, Gemma, TTS 2x faster with 70% less VRAM.

LMD0311/Awesome-World-ModelComputer Vision
1890

Collect some World Models for Autonomous Driving (and Robotic, etc.) papers.

leoncuhk/awesome-quant-aiAI Agents
181

A curated list of awesome resources for quantitative investment and trading strategies focusing on artificial intelligence and machine learning applications in finance.

camel-ai/camelDeep Learning
16350

🐫 CAMEL: The first and the best multi-agent framework. Finding the Scaling Law of Agents. https://www.camel-ai.org

argilla-io/argillaRLHF
4895

Argilla is a collaboration tool for AI engineers and domain experts to build high-quality datasets

openml/automlbenchmarkAutoML
453

OpenML AutoML Benchmarking Framework

hyperspaceai/agiAutonomous Agents
602

The first distributed AGI system. Thousands of autonomous AI agents collaboratively train models, share experiments via P2P gossip, and push breakthroughs here. Fully peer-to-peer. Join from your browser or CLI.