OpenClaw is a viral open-source AI agent project on GitHub, formerly known as Clawdbot and Moltbot.
Quick take: OpenClaw is the kind of project that’s equal parts “wow” and “please don’t run this on your work laptop first.” It’s powerful, but it wants grown-up security habits.
What it is
An open-source AI agent/assistant project on GitHub, reported as formerly named Clawdbot and Moltbot. Security researchers and press have flagged risks around exposed configs/tokens and infostealer-style credential theft; treat it like you’d treat a production credential store (because it effectively is one).
Gabriel’s notes
I saved this under Security & privacy because agent tooling is basically “automation with keys,” and keys are the whole ballgame.
Good fit if you want to:
- Run an agent locally and tinker with real workflows
- Learn how modern “computer-use” agents are wired
- Stress-test your own secret-management discipline
Pricing snapshot (auto-enriched):
The repo itself is free to access. Ongoing costs depend on what models/services you connect (Unknown / not confirmed).
Work-use / compliance snapshot (auto-enriched):
Assume it can touch sensitive data (tokens, API keys, files). Prefer sandboxing, least-privilege credentials, and no internet-exposed dashboards. License details: Unknown / not confirmed.
Alternatives (auto-enriched):
- AutoGPT — similar “agent” vibe; ecosystem differs (Unknown / not confirmed).
- OpenHands (OpenDevin) — more dev-task oriented; different UX (Unknown / not confirmed).
Before you adopt it:
- Run in an isolated environment; rotate any creds you ever paste.
- Inventory integrations and restrict scopes before enabling “skills.”
Sources
- https://github.com/openclaw/openclaw
- https://www.businessinsider.com/sam-altman-hires-openclaw-creator-peter-steinberger-personal-ai-agents-2026-2
- https://www.techradar.com/pro/security/openclaw-ai-agents-targeted-by-infostealer-malware-for-the-first-time
- https://www.wired.com/story/openclaw-banned-by-tech-companies-as-security-concerns-mount