<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>AI Agents | Daniel C Ferreira &lt;dcferreira></title><link>https://dcferreira.com/tag/ai-agents/</link><atom:link href="https://dcferreira.com/tag/ai-agents/index.xml" rel="self" type="application/rss+xml"/><description>AI Agents</description><generator>Wowchemy (https://wowchemy.com)</generator><language>en-us</language><lastBuildDate>Fri, 18 Sep 2026 00:00:00 +0000</lastBuildDate><image><url>https://dcferreira.com/media/icon_hu4876d3cb97a4ad0b3222b8f65edef0fe_436070_512x512_fill_lanczos_center_3.png</url><title>AI Agents</title><link>https://dcferreira.com/tag/ai-agents/</link></image><item><title>agent-pawl</title><link>https://dcferreira.com/project/agent-pawl/</link><pubDate>Fri, 18 Sep 2026 00:00:00 +0000</pubDate><guid>https://dcferreira.com/project/agent-pawl/</guid><description>&lt;p>&lt;code>pawl&lt;/code> ensures your AI coding agents follow the steps you want, in the order you want.&lt;/p>
&lt;p>You write a workflow as a YAML state machine, and &lt;code>pawl&lt;/code> drives the agent through it, not moving on until a step&amp;rsquo;s check passes.
Deterministic steps run as scripts, agentic steps are handed to subagents, and postconditions are checked by &lt;code>pawl&lt;/code> itself,
so the agent can&amp;rsquo;t simply declare the job done.&lt;/p>
&lt;p>It ships as a &lt;a href="https://claude.com/claude-code" target="_blank" rel="noopener">Claude Code&lt;/a> plugin with a bundled Go binary, a skill and enforcement hooks.
Check it out &lt;a href="https://github.com/dcferreira/agent-pawl" target="_blank" rel="noopener">on GitHub&lt;/a>.&lt;/p></description></item><item><title>Budgebell</title><link>https://dcferreira.com/project/budgebell/</link><pubDate>Tue, 15 Sep 2026 00:00:00 +0000</pubDate><guid>https://dcferreira.com/project/budgebell/</guid><description>&lt;p>An agent-first desktop app that nudges you through habits (movement drills first, general habits too), shows you what to do, and logs what you actually did.&lt;/p>
&lt;p>It&amp;rsquo;s designed to be driven by an AI agent rather than by hand: a bundled local &lt;a href="https://modelcontextprotocol.io" target="_blank" rel="noopener">MCP&lt;/a> server lets your agent add, update and
disable habits, decide what you get nudged about, and read back the log of what you did or skipped.&lt;/p>
&lt;p>Everything stays on your machine: habit definitions, media and logs, with no network calls, no telemetry and no cloud (the MCP server is local too).
Built with Tauri and Svelte, with &lt;a href="https://github.com/dcferreira/budgebell/releases" target="_blank" rel="noopener">releases&lt;/a> for macOS, Linux and Windows.&lt;/p></description></item></channel></rss>