Kiro Crew and Hermes Agent are the two most interesting open-source agent frameworks of 2026. Both keep agents running when you’re not there. Both learn from experience. But they take fundamentally different approaches.
Kiro Crew is a persistent workspace with coordinated multi-agent teams. Hermes Agent is a self-improving agent with a learning loop that runs on your own server. The choice depends on whether you want a team of agents or a single agent that gets smarter over time.
The core difference
Kiro Crew (AWS, August 2026) is a persistent agent workspace. Multiple agents coordinate on tasks, run on schedules, trigger from webhooks, and present results through purpose-built Apps. It’s designed for teams and complex workflows.
Hermes Agent (Nous Research, February 2026) is a self-improving agent. It creates skills from experience, improves them during use, persists knowledge across sessions, and talks to you via Telegram or Discord. It’s designed for individuals who want an agent that grows with them.
Specs comparison
| Spec | Kiro Crew | Hermes Agent |
|---|---|---|
| Developer | AWS (Kiro team) | Nous Research |
| Launch date | August 2026 | February 2026 |
| GitHub stars | New (growing) | 225,000+ |
| License | Open source | MIT |
| Architecture | Multi-agent workspace | Single agent with learning loop |
| Session model | Cross-session, persistent | Cross-session, persistent |
| Triggers | Cron, webhooks, manual | Cron, manual |
| Messaging | Slack, Telegram, Discord, WeCom | Telegram, Discord, Slack, WhatsApp, Signal |
| Learning | Memory, lessons, skills | Memory, lessons, skills, user modeling |
| Apps | Yes (purpose-built UIs) | No |
| Model support | Kiro CLI providers | Any (OpenRouter, OpenAI, custom) |
| Self-hosting | Yes | Yes |
| Origin | Internal Amazon (MeshClaw) | Nous Research |
Architecture
Kiro Crew orchestrates multiple agents through Agent Client Protocol (ACP). The main conversation coordinates specialized sub-agents that handle work in parallel. Apps provide purpose-built interfaces for specific workflows.
Hermes Agent runs as a single agent with a learning loop. It creates skills from experience, improves them during use, and builds a user model over time. Sub-agents are available for parallel work, but the architecture is single-agent-first.
Learning capabilities
Both agents learn from experience, but in different ways:
Kiro Crew:
- Memory carries context across sessions
- Corrections become durable lessons
- Repeated patterns become reusable skills
- Workspace-scoped lessons for project-specific guidance
Hermes Agent:
- All of the above, plus:
- User modeling (Honcho dialectic)
- FTS5 session search with LLM summarization
- Skills self-improve during use (not just creation)
- Cross-session recall
Hermes Agent has the deeper learning loop. Kiro Crew has better multi-agent coordination.
Messaging integration
Kiro Crew: Slack, Telegram, Discord, WeCom. Desktop app, web dashboard, TUI.
Hermes Agent: Telegram, Discord, Slack, WhatsApp, Signal, Email. CLI, messaging gateway.
Hermes Agent supports more platforms. Both support the major ones.
Multi-agent coordination
Kiro Crew: Designed for multi-agent work from the start. Coordinated crew of agents, isolated contexts, parallel workstreams, Apps for purpose-built interfaces.
Hermes Agent: Single-agent-first. Can spawn sub-agents for parallel work, but the architecture is designed around one agent getting smarter over time.
If you need a team of coordinated agents, Kiro Crew. If you want one agent that learns deeply, Hermes Agent.
Model support
Kiro Crew: Runs on the Kiro CLI. Supports the models Kiro supports (AWS Bedrock, Anthropic, OpenAI, etc.).
Hermes Agent: Works with any model. Nous Portal, OpenRouter, OpenAI, custom endpoints. Switch with one command. No lock-in.
Hermes Agent has broader model support. Kiro Crew is tied to the Kiro ecosystem.
When to use Kiro Crew
- You need coordinated multi-agent teams
- You want purpose-built Apps for specific workflows
- You’re in the AWS/Kiro ecosystem
- You need webhook-triggered automation
- You want enterprise-grade security (OS-level sandbox, audit logs)
When to use Hermes Agent
- You want a self-improving agent that learns deeply
- You need to run on your own server (data privacy)
- You want to talk to your agent via Telegram/WhatsApp/Signal
- You prefer model-agnostic (no provider lock-in)
- You want MIT license (most permissive)
My take
These are not direct competitors. They solve different problems.
Kiro Crew is for teams that need coordinated autonomous work. Multiple agents working on different aspects of a project, triggered by webhooks, presenting results through purpose-built Apps. It’s the “AI team” approach.
Hermes Agent is for individuals that want a personal agent that grows with them. One agent that learns your preferences, creates skills from experience, and talks to you wherever you are. It’s the “AI companion” approach.
For a solo developer who wants an agent that gets smarter over time, Hermes Agent. For a team that needs coordinated autonomous workflows, Kiro Crew. For both, run both.
FAQ
Which is more popular?
Hermes Agent has 225,000+ GitHub stars (launched February 2026). Kiro Crew just launched (August 2026) but had 39,000+ internal Amazon users before open-sourcing. Different metrics, both impressive.
Can I use both?
Yes. They solve different problems. Use Hermes Agent as your personal assistant and Kiro Crew for team workflows. No conflict.
Which has better learning capabilities?
Hermes Agent. The learning loop with user modeling, self-improving skills, and cross-session recall is deeper than Kiro Crew’s memory/lessons/skills approach.
Which has better multi-agent coordination?
Kiro Crew. It’s designed for coordinated multi-agent teams from the start. Hermes Agent can spawn sub-agents but is single-agent-first.
Which is easier to set up?
Hermes Agent. One-liner install, works with any model, talks via Telegram in minutes. Kiro Crew requires the Kiro CLI and is more complex to configure.
Which is more secure?
Both are secure. Kiro Crew has OS-level sandbox and signed audit logs. Hermes Agent runs on your own server with command approval. For enterprise security requirements, Kiro Crew has more built-in features.
Related Articles
- Kiro Crew Complete Guide
- Hermes Agent Complete Guide
- How to Run Hermes Agent Locally
- Hermes Agent vs CrewAI vs AutoGen
- OpenAI Agents SDK vs LangChain vs CrewAI