<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Routecraft Blog</title><link>https://routecraft.dev/blog/</link><atom:link href="https://routecraft.dev/feed.xml" rel="self" type="application/rss+xml"/><description>Write TypeScript capabilities that send emails, manage calendars, and automate work. Expose them to any AI agent via MCP. The code-first alternative to Make.com.</description><language>en</language><item><title>Anatomy of a team agent harness</title><link>https://routecraft.dev/blog/anatomy-of-a-team-agent-harness/</link><guid isPermaLink="true">https://routecraft.dev/blog/anatomy-of-a-team-agent-harness/</guid><pubDate>Thu, 23 Jul 2026 00:00:00 GMT</pubDate><description>A chat loop with tools is not a harness. An agent that works for a team needs four primitives most personal harnesses skip entirely: delegation, a learning loop, self-aware capability gaps, and multi-channel presence, plus three platform rules that keep the whole thing governable.</description><category>ai-agents</category><category>agent-harness</category><category>organizations</category><category>memory</category></item><item><title>A skills repository is not an automation platform</title><link>https://routecraft.dev/blog/beyond-the-skills-repository/</link><guid isPermaLink="true">https://routecraft.dev/blog/beyond-the-skills-repository/</guid><pubDate>Tue, 14 Jul 2026 00:00:00 GMT</pubDate><description>Every organisation is building a central repo of AI skills, prompts, and agent definitions right now. It is the right first step, and it has a ceiling nobody talks about. A skill is advice for an AI assistant; the doing still runs on personal access. A maturity ladder for what comes next.</description><category>ai-agents</category><category>skills</category><category>platform-engineering</category><category>mcp</category></item><item><title>AI agents are still single-player. Your organisation isn&apos;t.</title><link>https://routecraft.dev/blog/ai-agents-are-still-single-player/</link><guid isPermaLink="true">https://routecraft.dev/blog/ai-agents-are-still-single-player/</guid><pubDate>Tue, 30 Jun 2026 00:00:00 GMT</pubDate><description>The personal-assistant wave gave every employee an agent with their own access, their own memory, and their own chat window. That model cannot serve a team. The unit of organisational AI leverage is the governed capability, not the personal agent.</description><category>ai-agents</category><category>organizations</category><category>mcp</category><category>capabilities</category></item><item><title>Stop trusting your LLM to behave. Enforce it.</title><link>https://routecraft.dev/blog/stop-trusting-your-llm-to-behave/</link><guid isPermaLink="true">https://routecraft.dev/blog/stop-trusting-your-llm-to-behave/</guid><pubDate>Tue, 16 Jun 2026 00:00:00 GMT</pubDate><description>System prompts are requests, not rules. Human review, LLM judges, and guardrail frameworks all raise the cost of an attack, and none of them is a boundary. Banks did not solve embezzlement by hiring more honest tellers. Give the agent hands, not keys.</description><category>ai-agents</category><category>security</category><category>guardrails</category><category>llm</category></item><item><title>Your first MCP server in TypeScript with Routecraft</title><link>https://routecraft.dev/blog/your-first-mcp-server-in-typescript/</link><guid isPermaLink="true">https://routecraft.dev/blog/your-first-mcp-server-in-typescript/</guid><pubDate>Fri, 29 May 2026 00:00:00 GMT</pubDate><description>A ten-minute walkthrough from `bunx create-routecraft` to a working MCP server that Claude Desktop and Cursor can call. No auth, no infrastructure, no boilerplate. Just a TypeScript function exposed to your AI agent.</description><category>mcp</category><category>routecraft</category><category>typescript</category><category>claude-desktop</category><category>cursor</category></item></channel></rss>