Watch one agent free. Starter is observability for all 32 runtimes. Pro adds the governance layer: policy gates, evals, and waste detection. Enterprise is for procurement: audit, SSO and a customer-hosted server.
$9/node/month after trial · cancel anytime
也可以先从入门版试用开始，随时升级。
1 node = 1 OpenClaw machine · All prices in USD
| 功能 | 状态 | 适用范围与配置方式 |
|---|---|---|
| 单点登录（SAML、OIDC） | 已提供 | 适用于托管仪表盘，由 WorkOS 提供身份代理。在入驻时为您的组织配置，登录仅限您组织的邮箱域名。客户自托管服务器不支持此功能。 |
| 团队角色 | 已提供 | 托管仪表盘提供所有者和成员两种角色。更细分的角色（如独立的报告查看者和运维人员）已在规划中。 |
| SCIM 用户预配 | 规划中 | 尚未开发。目前由所有者在仪表盘中邀请和移除团队成员。 |
| 审计日志 | 已提供 | 每台机器上都有防篡改的哈希链记录。可使用 clawmetry verify-integrity 校验，并通过审计日志 API 读取。 |
| SIEM 导出 | 已提供 | 由每台机器以 CEF 或 JSON 格式通过 syslog（UDP、TCP 或 TLS）发送到您的收集器。需要 Enterprise 许可证。 |
| 客户自托管部署 | 已提供 | 基于已发布的 Docker Compose 部署方案，在您的基础设施上运行单租户服务器。许可证在本机验证，无需连接 ClawMetry。目前尚无 Helm chart。 |
| 数据位置 | 已提供 | 客户自托管时，数据保留在您的基础设施上。托管云运行于单一 Google Cloud 区域 europe-west1。 |
| 可用性与支持 | 不提供 | 目前不提供可用性 SLA。支持时间、联系人和升级流程将在您的试点或合同中以书面形式约定。 |
或发送邮件至 enterprise@clawmetry.com，我们将在 1 个工作日内回复。
| 免费版 | 入门版 | Pro | 企业版 | |
|---|---|---|---|---|
| 运行时 | ||||
| OpenClaw, NVIDIA NemoClaw, Goose & Qwen Code runtimes | ✓ | ✓ | ✓ | ✓ |
| Paid runtimes (28 of the 32): Claude Code, Cursor, GitHub Copilot, Codex, Gemini CLI, Muse Code, Lovable, Cline, OpenHands, OpenWorker, OpenExecutive, OpenCode, Aider, n8n, Devin, Antigravity, Grok Build, Grok Bot, Hermes, Pi, Deep Agents, Kimi CLI, DeepSeek Harness, Exo, NanoClaw, PicoClaw, QM, Replit Agent | · | ✓ | ✓ | ✓ |
| Custom runtime HTTP ingest API | · | · | ✓ | ✓ |
| 可观测性 | ||||
| Overview, Tool catalog, Sub-agents, Memory, Crons | ✓ | ✓ | ✓ | ✓ |
| Brain feed retention | 24h | 7d | 30d | custom |
| Token + cost retention | 7d | 30d | 90d | custom |
| Tracing, spans, waterfall | basic | ✓ | ✓ | ✓ |
| Cost attribution + model attribution | · | ✓ | ✓ | ✓ |
| Per-runtime health timeline | · | ✓ | ✓ | ✓ |
| 分析与智能 | ||||
| Per-run waste flags (runaway, cold cache, bloated context) | · | · | ✓ | ✓ |
| Per-run compare (A vs B with deltas) | · | · | ✓ | ✓ |
| Error triage (resolve / mute known errors) | · | · | ✓ | ✓ |
| Self-Evolve findings + Fix-with-AI | · | · | ✓ | ✓ |
| Asset registry (skills, prompts, workflows) | · | · | ✓ | ✓ |
| Eval suite (LLM-as-judge scoring) | · | · | ✓ | ✓ |
| Anomaly detection (spend + error spikes) | · | ✓ | ✓ | ✓ |
| Cost optimizer + model routing | · | ✓ | ✓ | ✓ |
| 治理与管控 | ||||
| NeMo Guardrails surface | ✓ | ✓ | ✓ | ✓ |
| Approval queue (block tool calls by policy) | · | ✓ | ✓ | ✓ |
| Budget limits + alerts | · | ✓ | ✓ | ✓ |
| Custom alert webhooks + PagerDuty + OpsGenie | · | · | ✓ | ✓ |
| Tool catalog policy + pre-execution gate | · | · | ✓ | ✓ |
| 集成 | ||||
| Channel adapters (Telegram, Slack, WhatsApp, Discord, ...) | 3 | all 21 | all 21 | all 21 |
| OTel export (Datadog, Grafana, Honeycomb) | · | · | ✓ | ✓ |
| SIEM export (Splunk, QRadar, Elastic) | · | · | · | ✓ |
| Cloud 与多节点 | ||||
| Number of nodes | 1 | unlimited | unlimited | unlimited |
| Cloud sync (E2E encrypted snapshot) | · | ✓ | ✓ | ✓ |
| Fleet view (multi-node dashboard) | · | ✓ | ✓ | ✓ |
| Self-hosted cloud dashboard | · | · | ✓ | ✓ |
| Offline licence on a customer-hosted server For air-gapped networks, the offline mode is documented; the full deployment on your network is validated in a pilot | · | · | · | ✓ |
| 安全与合规 | ||||
| E2E encryption (AES-256-GCM) | ✓ | ✓ | ✓ | ✓ |
| Secret redaction at ingest | ✓ | ✓ | ✓ | ✓ |
| Tamper-evident hash chain audit log | ✓ | ✓ | ✓ | ✓ |
| SSO (SAML, OIDC) on the hosted dashboard Not available on the customer-hosted server | · | · | · | ✓ |
| SCIM user provisioning | · | · | · | roadmap |
| Teams with owner and member roles Finer-grained RBAC and workspace scoping are on the roadmap, not shipped | · | · | · | ✓ |
| Data residency you control via self-hosted | · | · | · | ✓ |
| 支持 | ||||
| Community Discord + GitHub issues | ✓ | ✓ | ✓ | ✓ |
| Email support response | · | 24h | 4h | agreed in contract |
免费研究报告 / 2026 年 9 月
为什么可观测性至关重要。
了解可能出现的问题、需要关注的信号，以及何时介入。一份面向投入使用 AI 智能体团队的实用指南。
17 页英文 PDF。研究内容核查于 2026 年 9 月 17 日。
