OpenClaw SOUL.md: how to build your agent's personality
177 templates on GitHub, most users never touch theirs. How SOUL.md shapes every MCP tool call, the workspace file hierarchy, starter template, and five mistakes that waste tokens.
Read moreTutorials, deep dives, and insights on Openclaw and Model Context Protocols
177 templates on GitHub, most users never touch theirs. How SOUL.md shapes every MCP tool call, the workspace file hierarchy, starter template, and five mistakes that waste tokens.
Read more19.4M weekly npm downloads. Zod schemas for type-safe tool inputs, two API levels, Streamable HTTP transport, MCP Inspector testing, and deployment to Cloudflare Workers or Docker.
Read moreThree multi-agent patterns, real configs, MCP tool isolation across agents, memory boundaries, and community orchestration systems like SWAT and Lobster. When to scale beyond a single agent.
Read more9,080+ MCP servers on PulseMCP alone. We tested dozens across eight categories and ranked the ones worth installing. Developer tools, databases, web search, productivity, and more.
Read more8,590+ MCP servers available, three different clients to connect them to. Config files, transport types, Docker MCP Toolkit, and the 7 most common connection failures with fixes.
Read more250K GitHub stars in 60 days. But what does OpenClaw actually do? How it works, the MCP layer that makes it extensible, the ClawHavoc attack you should know about, and how to get started safely.
Read moreAn MCP server in Python takes about 15 lines of code. From FastMCP setup to tool definitions, testing with MCP Inspector, and connecting to Claude Desktop or OpenClaw. Full code included.
Read more13,700+ skills on ClawHub, 1,184 malicious ones found during ClawHavoc. How to browse, vet, install, and build your own OpenClaw skills with MCP server integration.
Read more42,000+ exposed instances, 1,184 malicious skills, 8 CVEs. Five real attacks targeting OpenClaw's MCP layer with the specific config fix for each.
Read more84.2% attack success rate. 5.5% of public MCP servers already poisoned. How tool poisoning works and a 3-layer defense playbook.
Read moreWith 1,184 malicious skills in ClawHub and a Docker sandbox bypass CVE, here's the hardening guide for your MCP connections. Covers real configs, network isolation, and per-server security.
Read more