MyClaw
One-click, 24/7 managed hosting that runs an open-source OpenClaw agent on a dedicated server for you — zero DevOps.
In brief
MyClaw is a managed agent platform: One-click, 24/7 managed hosting that runs an open-source OpenClaw agent on a dedicated server for you — zero DevOps. Hosting is managed cloud hosting on kubernetes — one statefulset plus a dedicated persistent volume per bot; a free tier is available to start; and it's unsupported on MCP. Best suited to teams that want the platform to host and operate the agent for them.
What it is
MyClaw (myclaw.ai) is a single-tenant managed-hosting wrapper that runs an open-source OpenClaw (or Hermes) agent on a dedicated, isolated per-user server — a managed Kubernetes StatefulSet with a persistent volume per bot. It packages install, config, updates, and backups behind a simplified web UI reachable over email, Telegram, WhatsApp, and browser, aimed at non-technical power users who want an OpenClaw agent without setting one up. It is explicitly independent of and unaffiliated with OpenClaw, and is bring-your-own-LLM-key: you pay for hosting, model tokens are billed to you separately by the provider.
At a glance
7/7 facts sourced- Pricing
- Hosting only, LLM tokens billed separately by you. Annual plans (default "Save 45%"): Lite $199/yr ($16.6/mo, reg. $29/mo), Pro $399/yr ($33.3/mo, reg. $59/mo), Max $799/yr ($66.6/mo, reg. $119/mo), Ultra $1,599/yr ($133+/mo, reg. $239/mo). No free tier — 7-day money-back only. Because model tokens are uncapped and billed by the provider, real all-in cost for an always-on agent is higher than the sticker; other MyClaw-branded surfaces publish different figures.source ↗
- Hosting model
- Managed cloud hosting on Kubernetes — one StatefulSet plus a dedicated persistent volume per bot, one bot per instance. "Stop" scales to zero while retaining data; independent scaling, hot config reload, and log streaming are offered. No self-host path and no published SLA are documented; data is on a 30-day PVC retention after which it is unrecoverable.source ↗
- Memory & state
- OpenClaw's built-in file memory persisted on the per-bot persistent volume. No managed vector store or knowledge graph is offered.source ↗
- MCP support
- Nonesource ↗
- Integrations
- Around six documented messaging channels — Discord, Telegram, Slack, and WhatsApp/Messenger/Instagram (the Meta ones via browser automation rather than official APIs). No voice or phone. Marketing cites "100+ integrations," but the roughly 20 tool connectors shown are OpenClaw's own ClawHub connectors, not a proprietary framework; channel credentials are entered manually rather than provisioned via per-agent OAuth.source ↗
- Model providers
- Bring-your-own-key across five providers, all pass-through: Anthropic, OpenAI, Google, Groq, and OpenRouter. Keys are AES-256-GCM encrypted, never returned to the browser, and rotatable without restart. There is no proxy, metering, per-tenant policy, credit pool, or fallback routing — no pooled or bundled model access.source ↗
- Open source
- Nosource ↗
Editorial scorecard
Overall 1.7 / 5
Our opinion — the same 0–5 rubric is applied to every platform, including Alfe. See the rubric.
Strengths
- Genuinely frictionless setup — a working hosted agent in roughly 30–60 seconds with no Docker, Python, ports, or terminal.
- Clean isolation primitive: one Kubernetes StatefulSet and a dedicated persistent volume per bot, with scale-to-zero "stop" that retains data, hot config reload, and log streaming.
- Honest bring-your-own-key economics for users who already hold model keys — you are charged only for hosting, and keys are AES-256-GCM encrypted and never returned to the browser.
- Stays on upstream, unforked OpenClaw, so users get the real runtime plus the ClawHub skills ecosystem; the operator also ships a genuine open-source content engine (openclaw-master-skills, openclaw-guardian).
- Multi-client pairing (browser, mobile, and CLI over WebSocket on a personal subdomain).
Trade-offs
- Reliability is the recurring complaint — reviews and a real paid-user report flag early stability issues, frequent server reboots, and support lag.
- No pooled or bundled LLM — bring-your-own-key only, across just five pass-through providers, so uncapped model-token spend is billed separately and all-in cost is unpredictable.
- Thin platform depth: no MCP or self-bootstrap, no managed memory service (only OpenClaw file memory), and no teams, projects, orgs, RBAC, or fleets — pricing is N bots on one personal account.
- Few first-class channels (about six), no voice or phone, and the Meta channels rely on fragile browser automation rather than official APIs.
- Corporate opacity and brand fragmentation — no named founders or legal entity surfaced, and several unrelated "MyClaw" vendors share the name with inconsistent pricing, making due diligence hard; despite "full code control" marketing, no hosting or orchestration code is open source.
Our take
Editorial opinion — not a hands-on test resultIn our opinion, MyClaw is a narrow but competent thing: a single-tenant "run OpenClaw on a box for you" hosting wrapper aimed at non-technical power users, not a multi-agent platform. It does the packaging genuinely well — setup is frictionless, the per-bot Kubernetes isolation is a sound design, the bring-your-own-key economics are transparent for those who already hold model keys, and the operator's open-source tooling (openclaw-guardian, openclaw-master-skills) is a real community strength. The trade-offs, on our reading, are depth and trust: no MCP, no managed vector or knowledge-graph memory, no teams or fleets, only around six channels, bring-your-own-key spend that is billed separately and hard to predict, a recurring reliability complaint in reviews, and a fragmented, corporately opaque brand shared with several unrelated vendors. Held to the same rubric as everything else here, it lands low-to-mid: fair on developer experience for its scope, but behind the managed leaders on memory, integrations, MCP, and scalability.
Sponsored
Alfe is a managed agent-hosting platform. If you want an agent hosted and running without wiring infrastructure yourself, see Alfe → (or read its dossier).
FAQ
- Does MyClaw include the LLM, or do I bring my own key?
- You bring your own key. MyClaw charges for hosting only — model tokens are billed to you separately by the provider (Anthropic, OpenAI, Google, Groq, or OpenRouter). There is no pooled or bundled model access and no credit pool.
- Does MyClaw support MCP?
- No. MyClaw does not offer native MCP or a self-bootstrap flow. It hosts an OpenClaw agent with OpenClaw's own file memory and ClawHub connectors.
- Is MyClaw open source?
- The hosted runtime is upstream open-source OpenClaw, but MyClaw's own hosting and orchestration control plane is closed — the LeoYeAI/MyClaw repository is a marketing/README-only project. "Full code control" refers to shell access to your own instance, not to the platform.
Compare MyClaw
Sources
- MyClaw homepage · accessed 2026-07-03
- MyClaw pricing · accessed 2026-07-03
- MyClaw OpenClaw hosting page · accessed 2026-07-03
- MyClaw docs (providers, channels, plans) · accessed 2026-07-03
- LeoYeAI/MyClaw GitHub repository · accessed 2026-07-03
The Agent Examiner editorial · last updated 2026-07-03. Facts are drawn from each vendor’s own public materials; scores are our editorial opinion (methodology).
Product and company names are trademarks of their respective owners and are used here only to identify the products under discussion.