Robotics & embodied
agents that plan, code and debug for robots and simulators
PhyAgentOS/PhyAgentOS PhyAgentOS is the clearest adoption leader, offering a session-centered runtime that lets the same agent workflows run unchanged across simulation and real hardware; if you need a single, well-adopted foundation for sim→real agent execution and multi-layer safety, start there. Choosing otherwise usually reflects a specific need the leader doesn't prioritize (ROS/MCP wiring, benchmarks, strict offline operation, specialized safety patterns, or out‑of‑the‑box GUIs).
Which one matches your setup?
Answer any of the questions — the shortlist updates as you go. Recommendations come from the capability passports below, nothing else.
Comparison matrix
Axes are extracted from each project's docs by our review pipeline; the maturity score is computed from stars, growth and commit activity — not an opinion. Click a column to sort.
| Runs in | Models | Context | Cost to run | |||||
|---|---|---|---|---|---|---|---|---|
⭐ 3.0k +48/7d | CLI | Fixed provider | Workspace & protocols | Multi-layer validation | ●●●●● | Self-hostable | ●●●●● | Free, self-hosted |
⭐ 3.0k +48/7d | CLI | Fixed provider | Workspace-wide context | Multi-layer safety | ●●●●● | Self-hostable | ●●●●● | Free, self-hosted; optional external services |
⭐ 3.0k +48/7d | CLI | Fixed provider | Workspace-wide protocol | Multi-layer safety | ●●●●● | Fully self-hosted | ●●●●● | Free, self-hosted (local compute) |
⭐ 1.5k +10/7d | CLI | BYOKOpenAIAnthropicGemini | Full robot context | Minimal guidance | ●●●●● | Self-hostable | ●●●●● | Depends on external model providers; uses your model client / API key. |
⭐ 230 +5/7d | CLICIWeb app | OpenAIAnthropicGemini | Diff only | Basic filtering | ●●●●● | Self-hostable | ●●●●● | Free, self-hostable; optional model API costs |
⭐ 191 +6/7d | CLIIDE | BYOK | Tool-only | Sim-safe defaults | ●●●●● | Runs fully local | ●●●●● | Free; local GPU required for some backends/training |
⭐ 1.7k +5/7d | CLI | BYOK | System-level context | None mentioned | ●●●●● | Self-hostable | ●●●●● | Your LLM/API key (your usage costs) |
⭐ 878 +16/7d | CLIWeb appCI | BYOKGeminiFixed provider | Diff only | None mentioned | ●●●●● | Self-hostable | ●●●●● | Your API key or local GPU |
⭐ 607 +5/7d | CLI | BYOK | Diff only | None | ●●●●● | Self-hostable | ●●●●● | Open-source, self-hosted; model API costs may apply |
⭐ 147 | CLI | Fixed providerOpenAI | Prompt + few-shot | Screener + verify | ●●●●● | Fully local | ●●●●● | Local model checkpoints to download; optional OpenAI API key for the GPT judge |
⭐ 141 | CLIWeb app | Gemini | Single-script context | No explicit controls | ●●●●● | Cloud API (your key) | ●●●●● | Requires cloud LLM access (your API key). |
⭐ 95 +5/7d | CLIIDECoding-agent plugin | BYOK | Runtime-only | Basic filters | ●●●●● | Fully local | ●●●●● | Free, self-hosted (Docker/local) |
Capability profiles
Six axes, 0–5 each. The shape tells you the strategy: a wide hexagon is a generalist, a spike is a specialist. Showing the 8 most established — the rest are in the full catalog.
Decouples cognition from physical hardware with a session-centered runtime and adapter/bridge architecture that lets the same sessions run unchanged across simulation and real robots while enforcing multi-layer safety and auditable execution.
Provides a session-centered runtime that decouples cognition from physical targets so the same agent workflows run across simulation and real hardware via small target adapters.
Decouples cognition from physical hardware with a session-centered runtime and small target adapters so the same agent codebase runs across sim and real robots.
Bridges ROS and the MCP standard to let LLMs fully observe and control robots without modifying robot source code.
Provides an auditable, fail-closed physical execution runtime that binds high-level agent intent to robot bodies and issues evidence-bearing receipts and leases.
Unifies simulation and real hardware under one natural-language Agent API with built-in mesh networking and ROS 2 interoperability, letting the same code run in sim or on robots.
Integrates LangChain LLMs with live ROS1/ROS2 systems to let developers inspect, diagnose, and operate robots via natural language.
Provides an end-to-end, robotics-focused framework that benchmarks and improves Code-as-Policy agents with interactive robot gym environments, multi-tier benchmarks, a training-free agent, and RL fine-tuning tools bundled together.
All repositories (20)
PhyAgentOS is a Recursive Self-Improving (RSI) physical agent operating system that enables agents to recursi…
PhyAgentOS is a Recursive Self-Improving (RSI) physical agent operating system that enables agents to recursi…
PhyAgentOS is a Recursive Self-Improving (RSI) physical agent operating system that enables agents to recursi…
PhyAgentOS is a Recursive Self-Improving (RSI) physical agent operating system that enables agents to recursi…
ROSA 🤖 is an AI Agent designed to interact with ROS1- and ROS2-based robotics systems using natural language…
Connect AI models like Claude & GPT with robots using MCP and ROS.
[ICML 2026] A Framework for Benchmarking and Improving Coding Agents for Robot Manipulation
RAI is a vendor agnostic agentic framework for Physical AI robotics, utilizing ROS 2 tools to perform complex…
A unified, agentic system for general-purpose robots, enabling multi-modal perception, mapping and localizatio…
Agent skills that make AI coding assistants write production-grade robotics software. ROS1, ROS2, design patte…
Seamlessly integrate state-of-the-art transformer models into robotics stacks
Physical AI runtime for embodied agents — governed action, verified experience, physical memory, and skill evo…
Control robots and physical hardware with natural language through Strands Agents.
Official codebase for Spotter: Let the Embodied Model Lead, and the VLM Reflect for It
🦾Make your robot autonomous with LLM agent. Set it up with the same ease as normal agents in CrewAI or Autoge…
ROS2 MCP: the MCP server that lets AI agents see, understand and operate ROS 2 robots. MPL-2.0.
EmbodiedAgents is a fully-loaded ROS2 based framework for creating interactive physical agents that can unders…
A general-purpose robotic agent framework based on LLMs. The LLM can independently reason, plan, and execute a…
Offline fault diagnosis assistant for the AgiBot X2 Ultra humanoid: drop in a log package, get a three-round r…