Hermes Agent
Nous Research
250.6k
A self-hostable personal agent that can plan, use tools, remember context, and delegate work to subagents.
Independent, source-checked AI agent directory
Find the right AI agent for your task
Compare open-source projects and commercial products by real use case, setup effort, cost, dependencies, and risk.
Popular
282 agents checkedSources linkedNo paid rankingsUpdated Oct 2, 2026
Browse by category and ownership
Compare the trade-offs
Open-source agents emphasize inspectability and control. Commercial agents emphasize managed setup and integrated workflows.
Refine
deepseek-harness
deepseek-ai
241.8k
DeepSeek Harness: Everything is a Plugin.
OpenCode
Anomaly
211.3k
An open-source coding agent available in the terminal, desktop app, and editor-oriented workflows.
AutoGPT
Significant Gravitas
187.7k
The MIT-licensed original standalone AutoGPT agent for planning and completing multi-step tasks with tools.
firecrawl
firecrawl
187.6k
The context API to search, scrape, and interact with the web at scale. 🔥
langflow
langflow-ai
155.4k
Langflow is a powerful tool for building and deploying AI-powered agents and workflows.
langchain
langchain-ai
147.4k
The agent engineering platform.
Codex
OpenAI
127.6k
OpenAI's coding agent for terminal, IDE, desktop, and cloud workflows, with local repository tools and sandbox controls.
Browser Use
Browser Use
117k
A Python agent and tooling layer for navigating websites, filling forms, extracting information, and completing browser tasks.
Pi Coding Agent
Earendil Works
111.3k
A compact terminal coding agent and TypeScript toolkit with multi-provider models, extensions, and optional isolation.
TradingAgents
Tauric Research
109.5k
A multi-agent financial research system that assigns analyst, researcher, trader, and risk roles to discuss market decisions.
Gemini CLI
107.2k
Google's open-source terminal agent for coding, research, file operations, and tool-based workflows with Gemini models.
Our editorial method
Fit beats hype.
A useful directory should narrow the decision, not inflate the list. Every profile answers the same three questions.
- 01
Can it finish a real task?
We start with the job and expected output, not the feature list.
- 02
What does setup really cost?
Accounts, API keys, skills, permissions, time, and money stay visible.
- 03
Can the claim be checked?
Official pages and repositories make the starting evidence traceable.