OpenClaw has 196,000+ GitHub stars, an OpenAI-backed foundation, and a skills marketplace with 13,729+ community modules — including several for autonomous prediction market trading. Four months ago it didn’t exist. Three months ago it had a different name. Two months ago it had two different names. Here’s the timeline and why it matters if you’re building prediction market agents.

The timeline

November 2025: Peter Steinberger, an Austrian developer, publishes ClawdBot — a lobster-themed open-source autonomous agent framework built in TypeScript/Node.js. It’s self-hosted, model-agnostic, and uses a composable skills architecture where capabilities are modular, installable plugins.

Late January 2026: ClawdBot goes viral. 100,000 GitHub stars in roughly eight weeks. Reports of Mac Mini sellouts at retail as users rush to set up local instances. TechCrunch, CNBC, Fortune, Forbes, WIRED, and Nature all cover the project.

January 27, 2026: Anthropic issues a trademark complaint over the “Clawd” similarity to their “Claude” AI. Steinberger renames to MoltBot — a reference to lobsters shedding their shell.

January 27-29, 2026: During the social media handle transition, handle snipers hijack the @clawdbot X account. They use it to promote a fraudulent CLAWD token on Solana that briefly hits a $16 million market cap before collapsing. A reminder that crypto-adjacent open-source projects attract predators.

January 30, 2026: The project settles on OpenClaw — combining open-source ethos with the lobster heritage.

February 14, 2026: Steinberger announces he is joining OpenAI. Sam Altman calls him “a genius with a lot of amazing ideas.” OpenClaw transitions to an independent open-source foundation with OpenAI support.

March 2026: 196,000+ GitHub stars. 13,729+ skills on ClawHub. The foundation governance model is operational.

Why the naming history matters

Two practical reasons:

Legacy search traffic. Many tutorials, GitHub repos, Stack Overflow answers, and forum posts still reference “ClawdBot” or “Clawdbot.” The getclawdbot.com domain still hosts a download page. If you’re searching for help with OpenClaw and find ClawdBot content, it’s the same project, same codebase, same skills architecture.

SEO for the space. “ClawdBot prediction market” and “ClawdBot Polymarket” are search terms with real volume and zero competition. Anyone building content around agent betting tools should capture both the OpenClaw and ClawdBot variants.

What matters for prediction market builders

Three takeaways from OpenClaw’s trajectory:

1. OpenAI backing signals stability

Agent frameworks rise and fall fast. Most never reach production adoption. OpenClaw’s transition to a foundation with OpenAI support means the framework is likely to have sustained development, corporate contributions, and a growing skills ecosystem for years. If you invest in building OpenClaw skills for prediction market trading, the platform risk is lower than most alternatives.

2. Composable skills are the right model for prediction markets

The Agent Betting Stack has four layers: Identity, Wallet, Trading, Intelligence. Each layer changes independently — you upgrade your LLM without rebuilding your order execution, or switch wallets without touching your analysis pipeline. OpenClaw’s composable skills map perfectly onto this model:

  • Layer 1 (Identity): SIWE skills, ENS resolution, Moltbook verification
  • Layer 2 (Wallet): Coinbase Agentic Wallet skills, Safe multi-sig skills
  • Layer 3 (Trading): PolyClaw (Polymarket), BankrBot (multi-platform), Solana CLI PM (Kalshi)
  • Layer 4 (Intelligence): Model-agnostic LLM brain, vector memory for market research

Detailed walkthrough in the OpenClaw Prediction Market Guide.

3. Security is a real, unsolved problem

The 196K stars come with scale risks. Cisco scanned 31,000 ClawHub skills and found 26% contained at least one vulnerability. The “ClawHavoc” supply chain attack uploaded 341 malicious skills that installed macOS Stealer malware. Over 21,000 exposed instances were found leaking API keys and chat history.

For prediction market agents — which hold wallet credentials and execute financial transactions — these aren’t theoretical risks. Audit every skill. Run in isolation. Never expose your instance to the public internet. The security hardening checklist is non-optional.

The competitive context

OpenClaw isn’t the only path to autonomous prediction market trading. The landscape includes:

  • Olas Polystrat — Purpose-built autonomous Polymarket agent, 361+ daily active agents, claims 55-65% success rates (detailed comparison)
  • Polymarket’s official agent framework — Developer-focused, 1.7K GitHub stars, maximum control for Polymarket-only bots
  • Polyseer — Multi-agent Bayesian analysis for Polymarket and Kalshi
  • CrewAI — Role-based multi-agent orchestration (general-purpose, not PM-native)

OpenClaw’s advantage is breadth. Polystrat’s advantage is depth. The composable agent tools guide covers when to use which.

Bottom line

OpenClaw went from zero to the dominant open-source agent framework in four months, surviving a trademark dispute, a token scam, and a founder departure. Its composable skills model is the most natural fit for the multi-layered prediction market agent stack. The prediction-market-specific skills (PolyClaw, BankrBot, Solana CLI PM) are functional today and growing.

If you’re building autonomous betting agents, OpenClaw belongs in your evaluation alongside purpose-built alternatives like Polystrat and the official Polymarket framework. The OpenClaw Prediction Market Guide has the full setup walkthrough.