Skip to content

AI Agents

Personal AI agents and agentic coding assistants — self-hosted, multi-channel, and self-improving systems that operate autonomously on behalf of a user.

This domain covers the emerging category of always-on AI agents that go beyond chat interfaces to execute tasks, manage workflows, and learn from experience. The space exploded in late 2025 when open-source projects demonstrated that a single agent could unify messaging platforms, execute shell commands, automate browsers, and improve its own skills over time.

Landscape (2026)

The AI agent space has settled into three philosophical camps:

Philosophy Representative Focus
Deep coding Claude Code, Codex, Cursor Terminal-native code generation, refactoring, and CI/CD
Universal automation OpenClaw, nanobot Cross-platform personal assistant across 24+ messaging channels
Self-improving Hermes Agent Persistent memory, autonomous skill creation, evolutionary optimization

Not Mutually Exclusive

Many users run Claude Code + OpenClaw (coding + personal automation) or Claude Code + Hermes (development + research). The agents complement rather than compete.

Key Projects

Core Agents

  • OpenClaw — Open-source personal AI assistant (360k+ GitHub stars). Multi-channel gateway connecting 24+ messaging platforms to any LLM. Created by Peter Steinberger, MIT licensed.
  • Hermes Agent — Self-improving AI agent by Nous Research (48k+ GitHub stars). Built-in learning loop that creates and evolves skills from experience. MIT licensed.

Claw Family Ecosystem

The "Claw" family is the ecosystem of projects built around or extending OpenClaw:

Project Maintainer Purpose
NemoClaw NVIDIA Secure OpenClaw runtime inside NVIDIA OpenShell with managed inference
claworc Community Multi-instance orchestration dashboard for OpenClaw
AutoResearchClaw AIMING Lab 23-stage autonomous research pipeline from idea to paper
openclaw-termux Community OpenClaw on Android via Flutter + built-in terminal
Paperclip PaperclipAI Enterprise management layer on top of OpenClaw
ClawTeam Community Multi-agent swarm coordination with OpenClaw as default agent
clawdbot Legacy Original name for OpenClaw before the Anthropic trademark dispute

Supporting Infrastructure

Project Maintainer Purpose
OpenViking ByteDance Context database for AI agents (targets OpenClaw)
OpenSpace HKU DS Lab One-command evolution for OpenClaw, Claude Code, Codex, Cursor
Lobster OpenClaw Typed workflow shell for deterministic multi-step pipelines

Comparisons

Sources

Questions

Open

  • How does OpenClaw's security posture evolve after the 9 CVEs disclosed in March 2026 — is the ClawHub skill marketplace trustworthy?
  • Will Hermes Agent's self-evolution approach (DSPy + GEPA) become the standard pattern for agent improvement, or is it a research curiosity?
  • Does NemoClaw's OpenShell sandboxing model scale for enterprise deployments, or is it still early-preview quality?
  • What is the long-term governance model for OpenClaw after Steinberger joined OpenAI and handed off to the non-profit foundation?
  • Can Lobster's workflow engine compete with LangGraph and similar DAG-based orchestration frameworks?

Answered

  • Q: What is the "Claw" family? — Ecosystem of projects extending OpenClaw: NemoClaw (NVIDIA security), claworc (orchestration), AutoResearchClaw (research), openclaw-termux (Android), Paperclip (enterprise), ClawTeam (swarm coordination).
  • Q: Is OpenClaw related to Claude/Anthropic? — Indirectly. Originally named "Clawd" after Anthropic's Claude, it was renamed after trademark complaints. No technical or organizational relationship.
  • Q: Which agent for coding vs personal automation? — Claude Code for deep coding, OpenClaw for universal messaging/automation, Hermes Agent for self-improving research workflows. Many users combine two.