Side-by-side
| Axis | Hermes Agent | ZeroClaw |
|---|---|---|
| Setup | One install script (install.sh), or its Docker image. | One install script or a prebuilt binary (Linux, macOS, Windows, every Raspberry Pi), then zeroclaw quickstart. |
| Security model | Commands run on the host by default, the dangerous ones behind approvals; Docker and cloud sandbox backends are opt-in. Chat users need an allowlist entry or pairing. | Supervised by default: medium-risk actions need approval, high-risk ones are blocked, file tools stay in the workspace and tools run in an OS sandbox it detects. Network egress stays open, and command execution is not audited. |
| Model support | Multi-LLM, all major providers. | Multi-LLM: Anthropic, OpenAI, Ollama, Bedrock, Gemini, Azure, OpenRouter and any OpenAI-compatible endpoint. |
| Cost | Free software; its README says it runs on a $5 VPS. | Free software (MIT or Apache-2.0); its docs say the runtime runs comfortably on any Raspberry Pi. |
| Ecosystem | Growing plugin ecosystem. | 30+ channels, MCP servers, skills and WebAssembly plugins; fewer GitHub stars than Hermes Agent (about 33,000 against 255,000 in October 2026). |
| Best for | Most users. Broad capability, with a container backend added for untrusted input. | Small hosts, hardware projects (GPIO, I2C, SPI) and setups that want approvals and a sandbox before the first run. |
Verdict
Hermes Agent for the widest set of tools and integrations, once you add a container backend. ZeroClaw if you'd rather start from a locked-down default in a single binary. Different philosophies; both legitimate.
Notes
- ZeroClaw is dual-licensed MIT or Apache-2.0. Until October 2026 this page called it BSL-licensed and local-only, with capabilities granted in signed YAML manifests; its docs describe a TOML config file and none of that.
- ZeroClaw's tool receipts sign successful tool calls with an HMAC, but its docs say they are not a durable audit log yet.
- Hermes Agent moves closer to ZeroClaw's default once you choose a container backend; both put risky commands behind an approval step.
Sources
- Hermes Agent README (v0.22.2)
- Hermes Agent docs: tools and terminal backends
- Hermes Agent docs: security
- ZeroClaw README (v0.8.6)
- ZeroClaw NOTICE: MIT or Apache-2.0
- ZeroClaw docs: security model
- ZeroClaw docs: sandboxing
- ZeroClaw docs: provider-agnostic
- ZeroClaw docs: tool receipts
Going deeper
For the full landscape report including hosting economics, security posture and regulatory context, see the 2026 landscape report. For the OpenClaw-specific history, see the complete OpenClaw timeline.
New comparison requests are welcome: subscribe and reply to any edition with your short-list.