Rick Boers vs OpenClaw
Explore what we can build. Examples describe possible workflows; integrations, hosting, and scope are agreed for your project.
OpenClaw is a free, MIT-licensed AI agent runtime you install on your own hardware (Mac, Windows, Linux, or a VPS) that connects LLMs of your choice to your files, shell, browser, and 20+ messaging channels, running proactively via a heartbeat daemon and cron jobs — but you supply the computer, the API keys, and the security posture. Launched in November 2025 by Peter Steinberger as Clawdbot (briefly Moltbot before becoming OpenClaw in late January 2026), it became the fastest-growing repository in GitHub history, reaching roughly 383,000 stars by July 2026. Since July 8, 2026 it has been stewarded by the OpenClaw Foundation, a 501(c)(3) chaired by Dave Morin with corporate sponsors including OpenAI, Microsoft, and Nvidia; Steinberger, who joined OpenAI in February 2026, retains technical leadership.
Rick Boers is an independent AI product builder. I design and build custom agents, document workflows, internal tools, and complete web and mobile applications, with one person owning the work from brief to delivery.
At a glance
| Comparison dimension | Rick Boers | OpenClaw |
|---|---|---|
| Always-on computer | Hosting and runtime architecture are chosen for your project; dedicated VMs are one option. | No hosted machine — you supply and operate the computer (spare Mac, $5–15/mo VPS, or per the official docs even a "Raspberry Pi-class box" with 4 GB RAM); it's only always-on if you keep your own hardware running. |
| Tool integrations | I build integrations around the tools your workflow needs. Access, permissions, API availability, and any service subscriptions are checked and scoped before implementation. | Full shell/file/browser access on its host plus community skills from the ClawHub marketplace (SKILL.md format, Claude Code/Cursor-compatible); skills are community-submitted, largely unvetted, and run unsandboxed, so quality and safety vary. |
| Messaging channels | Messaging or web interfaces are implemented where supported and included in the agreed scope. | 20+ channels — WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage, IRC, Microsoft Teams, Matrix, LINE, Mattermost, Nostr, Twitch, Zalo, WeChat, QQ and more — plus official iOS/Android/macOS apps and a local web Control UI; messages you proactively via its heartbeat. |
| Persistent memory | Storage and memory can be designed for the workflow, with retention and access controls agreed first. | Persistent memory as plain Markdown/YAML files (MEMORY.md, daily logs) on your own disk — transparent, greppable, git-versionable, but a file convention rather than a managed typed store. |
| Background automations | Scheduled and background jobs can be included with monitoring, limits, and human review where needed. | Background Gateway daemon with a configurable heartbeat (every 30 minutes by default) reading a HEARTBEAT.md checklist, plus cron jobs and webhooks; the July 2026 v2026.7.1 release added crons that react to command completion and watched conditions. |
| Pricing | The first 20-minute fit call is free. Paid work is scoped individually. We agree the deliverables, price, and payment milestones before work begins; hosting and model costs depend on the system we build. | Check the provider’s current pricing and plan limits. |
Where OpenClaw shines
- Free, MIT-licensed, and one of the largest open-source communities in AI: roughly 383k GitHub stars by July 2026, 532 contributors on the July 2026 (v2026.7.1) release alone, and since July 2026 a funded 501(c)(3) foundation with a full-time team of ten and sponsors including OpenAI, Microsoft, Nvidia, and GitHub.
- Complete data ownership: conversations and memory live as plain Markdown/YAML files on your own disk that you can grep, edit, and version with Git — nothing is locked in a vendor cloud.
- Broadest messaging-channel coverage in the category: 20+ channels including WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage, IRC, Microsoft Teams, Matrix, LINE, Mattermost, Nostr, Twitch, Zalo, WeChat, and QQ, plus official iOS, Android, and macOS apps.
- Deeply hackable: full shell and file access, a ClawHub skills marketplace using a SKILL.md format compatible with Claude Code and Cursor, remote Chrome browser control, and model-agnostic support including GPT-5.6, Claude Sonnet 5, Gemini, and local models via Ollama.
- No per-seat or per-agent fees — run as many agents as your hardware can handle and pay only raw API costs, or run local models for near-zero marginal cost.
Where it falls short
- There is no hosted computer: you must provision, patch, secure, and keep online your own machine or VPS, and the agent stops when that machine does.
- Security is your responsibility and has real failure history: CVE-2026-25253 (a CVSS 8.8 one-click RCE, patched in v2026.1.29 within days) and internet scans that found tens of thousands of exposed instances — 21,000+ in initial scans, 30,000+ by mid-February 2026 per Censys, Bitsight, and Hunt.io — many unauthenticated or unpatched.
- The skills ecosystem is unvetted: Koi Security's early-February 2026 audit found 341 malicious ClawHub skills (most from one campaign pushing a macOS infostealer), follow-up scans counted 800+ by mid-February, and Cisco's analysis of 31,000 agent skills found 26% contained at least one vulnerability — one of the project's own maintainers warns it is "far too dangerous" for users who can't run a command line.
- Setup and upkeep assume technical comfort: Node.js installs, environment configuration, API key management, and manual updates, versus no-code onboarding.
- There are no managed spend guardrails — raw API keys plus a default 30-minute heartbeat can quietly run up bills (the heartbeat alone can cost roughly $70/month on premium models; heavy users report $60–150/month and unoptimized setups have reported far more).
Choose OpenClaw if…
- You are technical, enjoy running your own infrastructure, and want full ownership of the agent's code, memory files, and data on hardware you control.
- Privacy or compliance requires everything on-prem — memory as local files and no vendor cloud in the loop is the whole point.
- You want to hack the agent itself: write custom skills, wire up unusual channels like Matrix, Nostr, IRC, or Zalo, or run local models via Ollama for near-zero marginal cost.
- You want zero software subscription and are happy trading your own setup and maintenance time for it.
Choose Rick Boers if…
- You need a custom workflow or product built around your team.
- You want one builder to own design, implementation, and delivery.
- Your project needs integrations or behavior beyond a packaged tool.
- You need a custom workflow or product built around your team.
Common questions
Is OpenClaw free to use?
The software is completely free and MIT-licensed with no subscription. You still pay for the LLM API calls it makes (2026 cost guides put moderate personal use around $10–30/month and heavier use with frequent heartbeats at $60–150+/month) and for the computer it runs on, whether that's a spare machine or a $5–15/month VPS. Local models via Ollama can cut API costs to near zero.
Is OpenClaw safe to run?
It's powerful but demands real security diligence: it runs with broad access to your files, shell, and accounts, skills aren't sandboxed, and early 2026 saw a CVSS 8.8 one-click RCE (CVE-2026-25253, patched within days), scans finding 21,000–30,000+ exposed instances, and hundreds of malicious skills uploaded to ClawHub. One of the project's own maintainers warns it's "far too dangerous" for users who can't run a command line. Run it on an isolated machine, audit skills, and set API spend limits.
Is OpenClaw better than Rick Boers?
This comparison is between using a packaged tool and commissioning a custom system. Provider features change; check current documentation. For a custom build, we agree the specific workflow, infrastructure, integrations, and operating responsibilities.