ClawTrust: review and alternative
A plain description of ClawTrust's managed OpenClaw hosting, the facts you need before buying, an honest review, and a side-by-side comparison with ControlClaw.
Updated November 7, 2026. ClawTrust details come from their public pages. Spotted something out of date? Tell us.
What is ClawTrust?
ClawTrust is a managed hosting service for OpenClaw, sold as an "AI employee". OpenClaw is the open-source, always-on AI agent that runs on its own machine, keeps memory between sessions and connects to email, chat and tools through skills. ClawTrust runs one OpenClaw agent for you on its own server, and you talk to it on Slack, Telegram, WhatsApp, Discord or email.
The company is ClawTrust, Inc., based in the United States, and it launched the hosted product in May 2026. Its pitch is security. The agent's server has no open ports and its disk is encrypted. API keys sit in a vault off the server, and a monitor on the host watches what the agent runs. Memory, called BrainTrust, stores what the agent learns and strips personal data before saving it.
In short: it is hardened OpenClaw hosting, one agent per plan, managed from a dashboard. That matters for the comparison. ControlClaw starts from the same idea, OpenClaw on its own server with secrets kept off it, and adds what ClawTrust does not have: screening of what the agent reads, approvals before risky actions, allow-lists, and a console for a fleet.
ClawTrust at a glance.
ClawTrust sells three monthly plans on its homepage: Lite at $29, Starter at $79 and Pro at $159. Each plan is one agent. Lite runs on your own free NVIDIA API key and skips web browsing; Starter adds browsing; Pro adds Python, smarter model routing and a $10 monthly AI allowance. Every plan has a hard AI budget cap ($5, $10 or $30 a month by tier); when it is used up, the agent pauses until the next month. There is a 5-day free trial, a card is taken at signup, billing is monthly only and fees are not refunded. Docs also mention an Enterprise plan at $299.
Each agent gets a dedicated virtual server from a cloud provider ClawTrust does not name. Its data processing agreement says data is processed mainly in the United States; EU servers are available on request. You pick a model (Claude, GPT, Gemini or 100+ others through OpenRouter), or bring your own key, or connect a Claude Pro or Max subscription. Support is by email. There is no public source code for ClawTrust's own layer, and no SOC 2 report yet.
- Type
- Managed OpenClaw hosting with a hardened server and a monitoring layer
- Agents per plan
- One; several agents can share one Slack workspace
- Where it runs
- A dedicated virtual server, provider not named; mainly United States, EU on request
- Models
- Claude, GPT, Gemini or 100+ through OpenRouter; bring your own key, or a Claude subscription
- Memory
- BrainTrust: kept by ClawTrust, personal data stripped, viewable and deletable from the dashboard; no export format published
- Setup
- About 5 minutes: name the agent, pay, pick channels
- Safety controls
- No open ports, encrypted disk, key vault off the server, tool policy engine, kernel monitoring (Pro and up), contact approval for chat DMs, hard AI spend cap
- Team admin
- A dashboard per agent; Slack routing for several agents; no seats or roles published
- Pricing
- Lite $29, Starter $79, Pro $159 a month; Enterprise $299; monthly only, 5-day trial with a card
- Open source
- OpenClaw is; ClawTrust's own layer is not published
ClawTrust review: a well-locked server, with the gate still open
We run the competing product, so read this with that in mind. We have tried to be fair; where ClawTrust is better, we say so.
As OpenClaw hosting, ClawTrust is careful in the right places. The agent's server listens on no ports and reaches the outside only through an outbound tunnel. The disk is encrypted. API keys, OAuth tokens and mail logins sit in a vault on ClawTrust's side. The agent sends its requests through a proxy that adds the key, and the proxy refuses any domain that key was not approved for. A policy engine checks every command the agent runs against known attacks. On Pro and up, a monitor in the server's kernel watches the agent from outside. The founder comes from security companies, and it shows.
The gap is what the agent reads, and what it is allowed to do. ClawTrust's pages describe no screening of incoming email, web pages or files before the agent sees them. A hidden instruction in a message can still steer it. The only approval step is pairing: a new person who DMs the agent on chat needs your okay once. There is no approval queue for sending an email, paying for something or changing a setting. There is no allow-list for who the agent may email, and no way to see several agents in one place. The policy engine is ClawTrust's, not yours; the pages do not show a way to set your own rules.
The commercial side is tighter than the marketing suggests. Each plan is one agent. Lite needs your own NVIDIA key and runs on shared hardware. The AI allowance is small and hard-capped, so a busy agent stops mid-month unless you bring your own key or subscription. Phone calls work through your own Vapi account, billed per minute by Vapi. Billing is monthly only, non-refundable, and the trial takes a card. Data is processed mainly in the United States; EU servers are on request.
What ClawTrust does well
- Server hardening is thorough: no open ports, encrypted disk, host-side monitoring
- API keys and logins kept off the agent, with a per-key list of allowed domains
- A hard monthly AI cap, so a bill cannot run away
- BrainTrust memory strips personal data and cleans itself up nightly on Pro and up
- Ready-made connections to 40+ business tools, including HubSpot, Salesforce and QuickBooks
Where it falls short
- No screening of incoming email, pages or files for hidden instructions
- No approval queue for email, payments or settings; only chat DM pairing
- No email allow-lists; no fleet console; one agent per plan
- Data mainly in the United States; EU on request, no region choice at signup
- Monthly only, no refunds, card for the trial; ClawTrust's own code is not published
Good for one person who wants a single OpenClaw agent on a locked-down server and trusts the vendor's rules. Not a fit for a team, for several agents, or for an agent that reads outside email and acts on it, because nothing checks that content or asks a person first.
Same engine. Same instinct on secrets. Different scope.
Both run the open-source OpenClaw on its own server, with secrets kept off it
Your skills, prompts and memory work on either. Both put API keys in a vault the agent cannot read.
Hardening versus governing
ClawTrust locks down the server and watches what runs on it. ControlClaw does that and also screens what the agent reads, asks a person before risky actions, enforces allow-lists, and shows every agent in one console.
One agent versus a fleet
ClawTrust fits one person with one agent. ControlClaw fits a team putting several agents on email, chat and calls.
Feature comparison: ControlClaw vs ClawTrust
| Feature | ControlClaw | ClawTrust |
|---|---|---|
| Basics | ||
| What you get | Managed OpenClaw agents plus AI Firewall, approvals and console | One hosted OpenClaw agent on a hardened server, with a dashboard |
| Open source | YesOpenClaw and ControlClaw, both on GitHub | PartialOpenClaw is open; ClawTrust's layer is not published |
| Setup | Sixty seconds from a template; channels and keys wired for you | About 5 minutes; you name the agent, pay and pick channels |
| Servers | Hetzner or AWS in EU, US, UK, Singapore, or your own cloud | A dedicated virtual server; provider not named, mainly US |
| Models | Any provider, or your own LLM subscription | Claude, GPT, Gemini or 100+ via OpenRouter; own key or Claude subscription |
| Safety | ||
| Secrets | YesHeld by the Firewall, never on the agent | YesVault on their control plane; a proxy adds the key per allowed domain |
| Prompt-injection screening | YesInbound mail, pages and files inspected first | NoNot described on their pages |
| Approvals | YesOne queue, full context, audit trail | PartialOne-time pairing for new chat contacts; no action approvals |
| Email and calendar allow-lists | YesBy domain, address or group | NoNot offered |
| Spend and tool limits | YesPer agent, enforced outside the agent | PartialHard AI cap per plan; tool policy is theirs, not yours |
| Capabilities | ||
| Meetings and calls | YesJoins Meet, Zoom, Teams and phone calls | PartialPhone calls through your own Vapi account; no meetings documented |
| Multiple agents | YesFleet, templates, shared Brain | PartialOne agent per plan; shared Slack routing, no fleet view |
| Repairs | One-click repair session you confirm | Health checks every 15 minutes, automatic restart |
| Backups | YesDaily and encrypted, included; kept 7 days and 4 weeks | UnknownNot documented |
| Commercial | ||
| Price | From $19/mo including a Small server | From $29/mo with your own NVIDIA key; $79/mo with browsing |
| Data location | EU, US, UK or Singapore, your pick | PartialMainly United States; EU on request |
| Trial | Fourteen days, no charge, card required | Five days, card taken at signup |
"Not described" and "Not documented" mean we could not find it on ClawTrust's site, docs or terms. If they publish it, we will update the row.
ClawTrust or ControlClaw: match the job.
Pick ClawTrust
You want one agent on a locked-down server.
- You are one person running one agent
- Server hardening matters more to you than approvals
- The agent mostly talks to people you have paired with it
- You are fine with data in the United States
- You want a fixed AI allowance that stops rather than bills
Pick ControlClaw
Your agents read outside email and act on it.
- Agents send email, join calls or hold passwords
- You want a human in the loop before risky actions
- You run more than one agent, or plan to
- More than one person needs to see what they do
- EU data location matters
ClawTrust vs ControlClaw pricing.
ControlClaw bills the plan flat and passes the server through at list price. ClawTrust bills one plan per agent, monthly only, with a small AI allowance that is capped.
ClawTrust vs ControlClaw: common questions.
Is ClawTrust the same as ControlClaw?
No. Both run OpenClaw on its own server and keep API keys off it. ClawTrust stops there: a hardened server and a monitor. ControlClaw adds an AI Firewall that screens what the agent reads, approvals before risky actions, allow-lists and a team console.
Does ClawTrust stop prompt injection?
Its pages do not say so. ClawTrust watches what the agent runs and blocks dangerous commands. It describes no check on the email, pages or files the agent reads. ControlClaw's Firewall inspects that content before the agent sees it.
Can I move my agent from ClawTrust to ControlClaw?
Yes. Export the workspace, memory and skills from OpenClaw and import them into a new agent. Channels are reconnected from the console. ClawTrust's terms keep a read-only export open for seven days after you cancel, so export first.
Which is cheaper, ClawTrust or ControlClaw?
ControlClaw Solo starts at $19/mo including the server, firewall and approvals. ClawTrust starts at $29/mo if you bring a free NVIDIA key and skip browsing, or $79/mo with browsing. Each extra agent on ClawTrust is another plan.
Where does my data live?
On ControlClaw, on a server in the EU, US, UK or Singapore, your pick. On ClawTrust, mainly in the United States; EU servers are available on request.
Sources (23)
- https://clawtrust.ai/
- https://clawtrust.ai/product
- https://clawtrust.ai/security
- https://clawtrust.ai/docs/security-architecture
- https://clawtrust.ai/docs/credentials
- https://clawtrust.ai/docs/data-privacy
- https://clawtrust.ai/dpa
- https://clawtrust.ai/docs/faq
- https://clawtrust.ai/docs/agent-capabilities
- https://clawtrust.ai/docs/infrastructure
- https://clawtrust.ai/docs/how-it-works
- https://clawtrust.ai/docs/model-routing
- https://clawtrust.ai/docs/claude-code-setup
- https://clawtrust.ai/docs/slack-integration
- https://clawtrust.ai/docs/channel-pairing
- https://clawtrust.ai/braintrust
- https://clawtrust.ai/blog/launch-technical-deep-dive
- https://clawtrust.ai/blog/openclaw-voice-agent-skills
- https://clawtrust.ai/blog/openclaw-free-trial
- https://clawtrust.ai/blog/clawtrust-vs-hostinger-openclaw
- https://clawtrust.ai/terms
- https://clawtrust.ai/llms.txt
- https://github.com/search?q=clawtrust&type=repositories