4 min

The open-source agent ecosystem

Hermes, OpenClaw, nanoclaw, aider, OpenHands, Goose, Continue, Cline.

Open-source autonomous agents are tools that wrap an LLM — usually any LLM — into a different shape than a chat or a CLI. Background daemons, multi-agent runtimes, IDE plugins, web sandboxes. You self-host them, give them an API key, and they go to work.

Hermes Agent

Nous Research

Background daemon. Persistent memory, scheduled tasks, Telegram/Discord.

github.com/NousResearch/hermes-agent

OpenClaw

Open source

Multi-agent runtime. Coordinates multiple coding CLIs as one.

docs.openclaw.ai

nanoclaw

nanoco.ai

Lightweight, containerised OpenClaw alternative. Messaging-app bridges (WhatsApp, Telegram, Slack), memory, scheduled jobs. Runs on Anthropic's Agent SDK.

github.com/nanocoai/nanoclaw

aider

Open source

Terminal pair programming. Mature, git-aware, multi-model.

aider.chat

OpenHands

All Hands AI

Autonomous engineer in a web UI. Local + customisable.

github.com/All-Hands-AI/OpenHands

Goose

Block

Extensible CLI/desktop agent. Strong tool ecosystem.

block.github.io/goose/

AutoGen

Microsoft

Multi-agent conversation framework. Python-first.

github.com/microsoft/autogen

crewAI

Role-based agent orchestration. Great for workflow automation.

www.crewai.com

Continue / Cline

Various

IDE-embedded copilots inside VS Code.

continue.dev

When to reach for these. When the polished products don't cover your case: background execution, cross-session memory, custom tool integrations, multi-agent orchestration, running fully on your own infrastructure, or swapping providers freely.

All of these are model-agnostic — you can point them at OpenAI, Anthropic, Google, Mistral, DeepSeek, or a local Ollama model. The next lesson explains how that wiring works.