KiloClaw

What Is KiloClaw? Hosted OpenClaw With Plan Questions

Kilo describes the product as a hosted OpenClaw service that creates a personal or organization-scoped agent without self-hosting. Each instance includes a web interface, Kilo Chat, persistent storage, model access through Kilo Gateway, and controls for starting, restarting, redeploying, and diagnosing the runtime.

The host includes a headless Chromium browser and an OpenClaw environment capable of files, shell work, web search, memory, messaging, schedules, and sub-agents according to the selected permissions and configuration.

Kilo's current product page says provisioning takes under several minutes. Older launch material used a shorter speed claim, so evergreen copy should use the current page rather than preserving a launch-day number.

KiloClaw Plans, BYOK Markup Claims & Trial Access

KiloClaw currently presents a managed-hosting subscription, with model inference billed separately through Kilo Gateway unless a free model or personal provider key is used. The offer also includes a limited amount of search-service credit.

Exact amounts are omitted because subscription fees, included services, and promotions can change. Check the live KiloClaw pricing section before comparing total cost.

Separate these cost layers:

• Managed VM hosting
• Pay-as-you-go model inference through Kilo Gateway
• Personal model-provider charges when using BYOK
• Search or research service usage
• Optional assisted setup
• Organization controls or support

The optional setup call shown on the site is not the hosting subscription. Likewise, a free model does not make the dedicated cloud host free.

Source note: This section summarizes public pages available at review time. A missing public workflow or architecture detail is treated as a public evidence gap to verify with the provider, not as proof that the feature is unavailable.

KiloClaw BYOK Flow, Channels & Setup Evidence

Kilo Chat is included without external token setup. The product and documentation also identify Telegram, Discord, and Slack, while the current product page lists email among its native integrations.

External services still require credentials, authorization, or pairing. The dashboard displays pairing requests when a new channel or device connects, allowing the user to approve the connection.

BYOK is optional and verified. Kilo Gateway is the default model route, but users can add supported personal provider keys. This creates three possible model-cost paths: gateway-billed models, available free models, or charges billed directly by the provider connected through a personal key.

Kilo says Gateway inference is routed at provider rates without a token markup. That statement applies to its Gateway billing, not to discounts, taxes, currency conversion, or charges a separate provider may impose on a BYOK account.

Source note: This section summarizes public pages available at review time. A missing public workflow or architecture detail is treated as a public evidence gap to verify with the provider, not as proof that the feature is unavailable.

KiloClaw Public Evidence to Verify Before Budgeting

This heading is incorrect under the current official offer: optional BYOK and a short trial are documented. Kilo also states that inference routed through its Gateway carries zero token markup, while personal-key usage is billed under the connected provider's own terms.

The more accurate limitations are operational:

• Hosting remains a paid subscription after the trial
• Model inference is separate from the host fee
• BYOK transfers model cost control and key management to the user
• Direct SSH is not provided; access is mediated through the agent and product interfaces
• A managed host does not guarantee compatibility with every OpenClaw skill
• Organization controls may require a different account or sales arrangement

The current page describes a brief evaluation period, not an unlimited free tier. Users should confirm what happens to the instance, storage, and integrations when the trial ends.

Source note: This section summarizes public pages available at review time. A missing public workflow or architecture detail is treated as a public evidence gap to verify with the provider, not as proof that the feature is unavailable.

KiloClaw vs MoClaw for Solopreneur Cloud Workflows

Both products provide managed cloud environments for agent work, but they package the experience differently.

KiloClaw exposes more infrastructure and agent-management concepts. MoClaw emphasizes assigning work through chat while its managed computer handles supported browsers, files, schedules, and skills.

Users should compare the current MoClaw integrations and pricing with Kilo's live offer rather than relying on model counts or launch promotions.

Security Questions: Keys, Sessions & Cloud Boundaries

KiloClaw has independently assessed infrastructure isolation, but that is not the same as saying every agent action is contained in a restrictive application sandbox.

The February 2026 security whitepaper describes dedicated Firecracker micro-VMs, per-user applications and networks, encrypted persistent volumes, encrypted credentials, authenticated routing, and a tested deletion process. The assessor reported no cross-tenant access path in the tested architecture.

The paper also records residual work: outbound internet access was unrestricted, additional egress controls were on the roadmap, and parts of supply-chain hardening were planned. It described explicit approval for shell execution at the time of assessment. Current product documentation says new instances use a broad tool profile by default, so users should verify the live approval policy rather than assuming a dated control still applies unchanged.

MoClaw states that agent work runs sandboxed in a dedicated cloud environment, separate from the user's personal device. It supports browser control, files, schedules, and skills inside that managed computer.

Local OpenClaw gives the operator direct infrastructure control, but enabled tools may reach local files, commands, devices, and authenticated browser sessions. In all models, limit credentials and permissions to the tasks the agent actually needs.

Source note: This section summarizes public pages available at review time. A missing public workflow or architecture detail is treated as a public evidence gap to verify with the provider, not as proof that the feature is unavailable.

Questions