ClawPanel
A cross-platform control panel for installing, configuring, diagnosing, and operating OpenClaw and Hermes Agent.
Per-dimension scores and reasoning
Evidence shows: README mentions AI assistant with 8 tools and 4 modes, and test files show media generation tools (generate_image, create_video_task) are marked as DANGEROUS_TOOLS and forced confirmation, independent of confirmDanger switch, indicating user confirmation mechanism. However, there is no systematic description of least privilege, such as Tauri permission configuration. Data flow transparency is insufficient; no explanation of how API keys are stored and transmitted. Sensitive data handling only mentions API keys stored in local openclaw.json, but no encryption measures. Dependency security: package.json lists dependencies but no vulnerability scanning or lock file evidence. External effects: README mentions one-click install script and Docker commands, but no security review. Rollback mechanism not mentioned. Source attribution: README and LICENSE clarify copyright and trademarks, but no detailed contributor guide. Deductions: least privilege, data flow transparency, sensitive data handling, dependency security, external effects, rollback lack sufficient evidence.
Evidence shows: README feature list aligns with test files covering channels (Telegram, WhatsApp, ClickClack, Nextcloud Talk, Twitch, Nostr, IRC), showing good self-consistency. Dependency availability: package.json lists dependencies but no lock file or version pinning evidence, and Docker install uses npm install instead of npm ci, introducing uncertainty. Failure messages: test files show diagnostic functions (buildOpenClawChannelDiagnosis) check credentials and return error details, but no documentation on how UI presents these errors. Deductions: dependency availability lacks lock evidence, failure messages only in tests, not documented.
Evidence shows: README targets beginners and experts, provides desktop and web versions, supports multiple platforms (Windows, macOS, Linux) and embedded devices (Orange Pi, Raspberry Pi), covering broad scenarios. Capability boundaries: README lists features but does not clearly define AI assistant's specific capability boundaries and limitations. Trigger precision: test files show media tools are disabled by default and require explicit enabling, but no description of other tools' trigger conditions. Environment fit: README provides Node.js version requirements (18+, recommended 22.19.0+) and Docker deployment, but no detailed Tauri environment requirements. Deductions: capability boundaries and trigger precision insufficiently described.
Evidence shows: README has clear structure with sections for features, installation, build, related projects, indicating good information architecture. Install notes are detailed, providing download center, Docker, source build options. Naming stability: project name and version are clear in package.json, but no naming convention documentation. Examples and FAQ: README provides quick start steps, but no FAQ or common issues. Known limitations: README does not explicitly list known limitations. License: LICENSE file contains AGPL-3.0 and additional terms, package.json license field is AGPL-3.0, README badge shows AGPL-3.0, consistent. Versioning and changelog: package.json has version, but no CHANGELOG file. Maintenance responsibility: README provides contact email and support, but no explicit maintainer identity. Deductions: missing FAQ, known limitations, CHANGELOG, and explicit maintainer info.
Evidence shows: Output usability: README describes rich features but no actual output examples or screenshots (except showcase image). Marginal value: project provides multi-engine management and AI assistant, offering differentiated value. Cost-benefit: project is open source, but no performance or resource usage data. Deductions: output usability lacks concrete examples, cost-benefit lacks data support.
Evidence shows: README feature claims partially correspond to test files, but no traceable links or evidence. Cross-source corroboration: README and test files corroborate each other, but lack third-party verification. Fact-inference separation: README claims are mostly assertions, not distinguishing facts and inferences. Deductions: claims lack traceability, cross-source corroboration limited, fact-inference not clearly separated.
- API key storage and transmission are not clearly specified; review permissions and encryption of openclaw.json.
- Dependencies are not version-locked; Docker install uses npm install, potentially introducing supply chain risks.
- No rollback mechanism provided; upgrade failures may not be recoverable.
- Media generation tools are force-confirmed, but confirmation mechanisms for other tools (e.g., shell) are not described.
What does this agent do, and when should you use it?
ClawPanel is a visual management panel for OpenClaw and Hermes Agent, available as a desktop application and a headless web deployment. The desktop build uses Vanilla JS, Vite, Rust, and Tauri v2, while web mode uses scripts/dev-api.js to invoke the local OpenClaw CLI. It covers service control, model and gateway settings, messaging channels, agents, memory, logs, scheduled jobs, and usage, alongside streaming chat and image recognition. Its built-in AI assistant offers Chat, Plan, Execute, and Unlimited modes to help inspect or modify a system under configurable terminal and file permissions. It fits individuals and teams operating either supported engine on a workstation or Linux server.
The panel starts and stops OpenClaw / Hermes Gateway, checks versions and upgrades, manages model providers, Base URL values and API keys, tests connectivity and latency, and configures ports, access scope, authentication tokens, and Tailscale. Its chat UI provides WebSocket streaming, Markdown rendering, and multimodal image input; Hermes views expose conversations, message flow, long-term memory, and tool-call details. The AI assistant uses eight named tools—ask_user, get_system_info, run_command, read_file, write_file, list_directory, list_processes, and check_port—to inspect configuration and system state, run shell commands, and write files when the selected mode permits it. Production web mode runs after npm run build with npm run serve and listens on 0.0.0.0:1420 by default.
- An OpenClaw operator can inspect Gateway state, ports, and logs, then restore a configuration backup from the service-management area.
- A team evaluating both Hermes Agent and OpenClaw can switch between the two engines and administer them from one panel.
- An administrator running an agent on an ARM64 board or headless Linux server can use web mode for browser-based remote administration.
- A user with several model providers can enter API keys and Base URLs, batch-test connectivity, and configure primary and fallback models.
- An administrator connecting agents to Telegram, Discord, Lark, DingTalk, or QQ can configure channels and bind multiple agents on the same platform.
What are this agent's strengths and limitations?
- It manages both OpenClaw and Hermes Agent, reducing the need to operate separate administration interfaces.
- It offers both a Tauri v2 desktop application and a Node.js web service; the documented web path covers headless Linux, Docker, and ARM64 environments.
- Its AI assistant combines eight concrete system, filesystem, process, port, and shell tools with four permission-oriented operating modes.
- Model configuration, gateway control, logs, memory, channels, scheduled tasks, and usage reporting are available in one product.
- Its core function depends on OpenClaw or Hermes Agent; it is not evidenced as a standalone general-purpose agent runtime.
- Web mode listens on all network interfaces by default; the documentation calls for an access password, firewall restrictions, and HTTPS reverse proxying, adding operational work for public deployments.
- Model use requires user-supplied provider API keys, so connectivity, account balance, and network conditions affect testing and chat.
- Desktop source builds additionally need Rust, Tauri v2 prerequisites, and Git; the Node.js version used for Gateway operation must meet the current OpenClaw requirement.
- License evidence is inconsistent: repository metadata says NOASSERTION while the README states AGPL-3.0.
How do you install or deploy this agent?
Desktop source builds require Node.js 18+, Rust stable, and Tauri v2 system prerequisites:
git clone https://github.com/qingchencloud/clawpanel.git
cd clawpanel
npm installRun ./scripts/dev.sh on macOS/Linux, or npm run tauri dev on Windows.
For Linux web deployment, run:
curl -fsSL https://raw.githubusercontent.com/qingchencloud/clawpanel/main/scripts/linux-deploy.sh | bashThen open http://SERVER_IP:1420. The repository also documents Docker and ARM64 web deployment paths.
How do you use this agent?
On first launch, complete the Node.js, Git, and OpenClaw environment checks and installation. In Model Configuration, add a provider’s Base URL and API key, use Test Connection, and save; the documented options include an OpenAI provider and an OpenAI Compatible provider. Start the Gateway in Service Management, then choose a model in Real-time Chat and send a message. For troubleshooting, use the AI assistant’s Plan mode to inspect configuration and logs, or Execute mode to run confirmed diagnostic and repair actions.