Side-by-side
| Axis | IronClaw | Hermes Agent |
|---|---|---|
| Setup | Release installers for macOS, Linux and Windows, then ironclaw onboard: provider, key, WebUI token and the background service. | One install script, then hermes setup. |
| Security model | Untrusted tools run in WebAssembly sandboxes limited to the hosts, paths and credentials they declare. Secrets are encrypted (AES-256-GCM) and injected at the host boundary, with a leak detector. The shell tool is off unless you set ALLOW_LOCAL_TOOLS=true, and then runs on the host. | Commands run on the host by default, with approvals for dangerous ones; Docker and cloud sandbox backends are opt-in, and its security policy asks for the whole agent in a container for untrusted input. |
| Model support | Multi-LLM: NEAR AI by default, Anthropic, OpenAI, GitHub Copilot, Ollama and OpenAI-compatible servers. | Multi-LLM: Nous Portal, OpenRouter, Anthropic, OpenAI, Gemini and local servers such as Ollama. |
| Cost | Free software (MIT or Apache-2.0). Model costs separate. | Free software; its README says it runs on a $5 VPS. Model costs separate. |
| Ecosystem | WebAssembly tools and channels installed as extensions, plus MCP servers. | Skills, plugins and MCP servers. |
| Best for | Setups that want untrusted tools sandboxed and secrets kept away from them from day one. | Setups that want broad tools on the host with approvals, and will add a container backend themselves. |
Verdict
Same licence family, different starting points. IronClaw starts locked down and asks you to grant tools what they need; Hermes Agent starts open on the host and asks you to add the sandbox. Pick the one whose default matches the input your agent will read.
Notes
- Until October 2026 this page priced IronClaw at $750 a seat, called it source-available and credited it with gVisor, RBAC, SAML SSO and an air-gap mode. Its repository, read on 2026-10-18, shows an MIT or Apache-2.0 licence and documents none of those.
- IronClaw's README lists a full audit log of tool executions; Hermes Agent writes logs under ~/.hermes/logs/ with secrets redacted.
- Both publish security documentation: IronClaw a security page and its README, Hermes Agent its security docs and SECURITY.md.
Sources
- IronClaw README (main, b0b999d): install, security, licence
- IronClaw docs: sandboxed WASM tools
- IronClaw docs: shell tool
- IronClaw .env.example: LLM backends
- Hermes Agent README (v0.22.2)
- Hermes Agent docs: security
- Hermes Agent SECURITY.md (v0.22.2): trust model
- Hermes Agent docs: AI providers
- Hermes Agent docs: configuration (directory structure, logs)
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.