Agent Support
Clawrium supports multiple agent types, each designed for different use cases. This section provides detailed support matrices for each agent.
Available Agents
Production Ready
| Agent | Description | Status |
|---|---|---|
| OpenClaw | Full-featured agent with multi-channel support | ✅ Production Ready |
In Development
| Agent | Description | Status |
|---|---|---|
| Hermes | Nous Research self-improving agent — local OpenAI-compatible HTTP API, file-based memory | 🚧 In Development |
| ZeroClaw | Minimal CLI-only agent for simple automation | 🚧 In Development |
Install-only CLI
| Agent | Description | Status |
|---|---|---|
| Claude Code | Isolated per-agent Claude Code environment with on-demand CLI and GUI chat; no daemon or native web UI | ✅ Supported contract |
Legend
| Symbol | Meaning |
|---|---|
| ✅ | Fully supported and tested |
| 🚧 | In development / Planned |
| ❌ | Not supported |
| 📋 | Not planned / Deferred (PRs welcome) |
Quick Comparison
| Aspect | OpenClaw | Hermes | ZeroClaw | Claude Code |
|---|---|---|---|---|
| Status | ✅ Production Ready | 🚧 In Development | 🚧 In Development | ✅ Install-only |
| Transport | Native daemon | Local OpenAI-compatible HTTP API (127.0.0.1:8642) | CLI process | Per-agent CLI environment |
clawctl agent chat <name> support | ✅ | ✅ (OpenAI-compatible HTTP backend) | 🚧 | ✅ (finite CLI backend) |
| Multi-Provider | ✅ (OpenAI, Anthropic, OpenRouter, Bedrock, Vertex, ZAI, Ollama) | ✅ (OpenRouter, Anthropic, OpenAI, Ollama / custom) | 🚧 (OpenAI, Anthropic, Ollama planned) | ❌ — one OAuth or Anthropic API-key credential |
| Memory model | Daily files + identity files | Two fixed files: MEMORY.md (≤ 2200 chars), USER.md (≤ 1375 chars) | ❌ | ❌ |
| Identity management | clawctl-managed SOUL.md / IDENTITY.md | Hermes-managed SOUL.md / AGENTS.md inside ~/.hermes/ (accessible via clawctl agent memory) | ❌ | Project-owned .claude settings untouched |
| Messaging gateways | Discord ✅, Slack 🚧, Web 🚧 | Discord ✅, Slack/Telegram/WhatsApp/Signal/email/... 📋 deferred | ❌ | ❌ |
| External integrations | GitHub 🚧, Jira 🚧 | 📋 Deferred | ❌ | ❌ |
| Onboarding wizard | ✅ 4-stage | ✅ 4-stage (identity auto-skipped) | 🚧 2-stage | ❌ — bounded configure/sync only |
| Resource usage | Moderate | Moderate-to-high (uv venv + npm + playwright) | Low | Low until an operator runs a command |
Choosing an Agent
Use OpenClaw when:
- You need Slack or multi-channel (Discord + Slack + Web) support
- You want clawctl-managed customizable identity/personality
- You need integrations with external tools
- You want a fully production-ready experience today
Use Hermes when:
- You want a local OpenAI-compatible HTTP API in front of your model
- You want Discord support backed by a local inference endpoint
- You want a self-managed-identity agent (hermes manages its own
SOUL.md/AGENTS.md) - You're driving a local inference endpoint (Ollama, vLLM, llama.cpp) and want hermes to wrap it
Use ZeroClaw when:
- You only need CLI interaction
- You want minimal resource usage
- You need quick automation scripts
- You prefer simple, no-frills setup
Use Claude Code when:
- You need an isolated, pinned Claude Code installation on a fleet host
- You will run finite native commands through
clawctl agent exec <name> -- <args...>(or a shell expression throughagent shell) or chat on demand throughclawctl agent chat <name>or the GUI Chat tab - You do not need a Clawrium-managed gateway, daemon, or native web UI
- You need Clawrium to own only the per-agent account, bounded global settings, selected credential environment, and installation prefix
Adding New Agents
To request support for a new agent type or feature:
- Check the existing agent matrices for current capabilities
- Open an issue describing your use case
- Reference the relevant agent matrix in your request
See CONTRIBUTING.md for contribution guidelines.