Hermes AgentvsQwenPaw
Generated from the two spec rows. Green marks the better value where a spec has a clear direction. Everything else is just different.
Hermes Agent
Nous Research · Agent harness
OSSMCP
Panel
6.84 spec wins
- Reliability
- 6.2
- Usefulness
- 6.5
- Cost
- 7.2
- Longevity
- 7.2
“Reachable on Signal and by email, so you can now be left on read by an agent on six platforms.”
QwenPaw
AgentScope (Alibaba) · Agent harness
OSSMCP
Panel
6.81 spec wins
- Reliability
- 6.3
- Usefulness
- 6.7
- Cost
- 7.3
- Longevity
- 6.7
“The cloud quick start includes a reminder to set the Studio to non-public, so strangers cannot control your assistant.”
Spec by spec
| Spec | Hermes Agent | QwenPaw |
|---|---|---|
| Architecture | ||
| Category | Agent harness | Agent harness |
| Runs | local, cloud, sandbox | local, cloud |
| Platforms | macos, linux, windows | macos, linux, windows |
| Context window | not documented | not documented |
| Protocols | ||
| MCP client | Yes | Yes |
| MCP server | Yes | No |
| Capabilities | ||
| Runs terminal commands | Yes | Yes |
| Multi-file edits | Yes | No |
| Git operations | No | No |
| Browser control | Yes | Yes |
| Sandboxed execution | Yes | Yes |
| Multi-agent orchestration | Yes | Yes |
| Headless / CI mode | Yes | No |
| Models | ||
| Backbone | Nous Portal, OpenRouter, OpenAI, Anthropic, any OpenAI-compatible endpoint, Ollama, vLLM, llama.cpp | Qwen, OpenAI, Anthropic, Gemini, DeepSeek, Kimi, OpenRouter, QwenPaw-Flash, Ollama, LM Studio |
| Bring your own model | Yes | Yes |
| Local models | Yes | Yes |
| Cost | ||
| Pricing model | byok | byok |
| Starts at | $20/mo | $0/mo |
| Free tier | Yes | Yes |
| Bring your own key | Yes | Yes |
| Openness | ||
| Open source | Yes | Yes |
| License | MIT | Apache-2.0 |
| GitHub stars | 250,643 | 35,413 |
Which one would each critic pick
| Critic | Hermes Agent | QwenPaw | Pick |
|---|---|---|---|
| El Juez | — | — | not enough reviews |
| El Amigo | 7.0 | 7.3 | QwenPaw — Pick QwenPaw if you want a personal agent in DingTalk, Lark or iMessage that runs on its own small models without a key; pick NanoClaw if you would rather have Docker walls and Claude. |
| El Crítico | 6.3 | 6.0 | Hermes Agent — An agent that writes its own skills from experience ships behaviour nobody reviewed, and the first place it runs them is your own shell. |
| El Profesor | 6.8 | 6.5 | Hermes Agent — Cross-session recall is FTS5 search over past sessions plus model summarisation, execution spans seven backends from local shell to Modal, and no benchmark is published. |
| La Inversora | 6.8 | 6.8 | no preference |
| La Jefa | 5.3 | 6.0 | QwenPaw — Kernel-level sandboxing on all three desktop OSes is the strongest security page on this board, a multi-user Hub shipped today, SSO is absent, and init --defaults accepts telemetry silently. |
| El Hacker | 8.5 | 8.0 | Hermes Agent — MIT, ~/.hermes with hermes config set, any OpenAI-compatible endpoint, Ollama, vLLM and llama.cpp, and MCP servers configured from the docs page; I can run this air-gapped. |
Picks are derived from each critic's own scores. Humans vote on matchups on the duels page.