{"id":34251,"date":"2026-06-25T17:39:58","date_gmt":"2026-06-25T15:39:58","guid":{"rendered":"https:\/\/www.angulararchitects.io\/blog\/ae-harness-setup-for-angular\/"},"modified":"2026-06-25T18:11:31","modified_gmt":"2026-06-25T16:11:31","slug":"ae-harness-setup-for-angular","status":"publish","type":"post","link":"https:\/\/www.angulararchitects.io\/en\/blog\/ae-harness-setup-for-angular\/","title":{"rendered":"Agentic Engineering: How to Set Up Your Harness for Angular Development"},"content":{"rendered":"
\n\t\t\t\n\t\n\t\n\t\t

\n\t\t\tThis is post 7 of 7 in the series “Agentic Engineering”<\/em>\t\t<\/p>\n\t\t\t<\/label>\n\n\t\t\t

\n\t\t\t
    \n\t\t\t\t\t\t\t\t\t
  1. Agentic Engineering: Which LLM Is Best for Angular Development?<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  2. Agentic: Which App\/Harness Is Best for Angular Development?<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  3. Agentic Engineering: What Does AI Coding Really Cost?<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  4. Agentic Engineering: What Do AI Coding Tools Do With Your Code?<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  5. Agentic Verdict: What’s the Best Solution for You?<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  6. Anthropic Just Released the First Next-Gen Model to the Public<\/a><\/li>\n\t\t\t\t\t\t\t\t\t
  7. Agentic Engineering: How to Set Up Your Harness for Angular Development<\/span><\/li>\n\t\t\t\t\t\t\t<\/ol>\n\t\t<\/div>\n\t<\/div>\n

    Over the last few weeks I've written about which LLM is best for Angular<\/em> development<\/a> and, right after, about which app or harness<\/a> to use around it. In that second post I treated the words app<\/em> and harness<\/em> as more or less the same thing and never really drew a line between them. This post is the follow-up that does exactly that.<\/p>\n

    It zooms in on the harness<\/strong>: what it actually is, and how to set one up for your Angular<\/em> projects so the agent not only runs smoothly, but produces generated and refactored code that I would actually accept in long-lived, continuously evolving Angular<\/em> codebases.<\/p>\n

    So What Is the Harness?<\/h2>\n

    Small naming note for Angular<\/em> readers: I do not<\/strong> mean the Angular CDK ComponentHarness<\/code><\/a> for testing components. In this post, a harness is the software around the model.<\/p>\n

    The harness is the thing that contains the model. And here's what a lot of people miss: on its own, even a strong LLM is basically just fancy autocomplete, like the one we've been using for years from GitHub Copilot. The only thing it can really do is respond to our messages by writing text. So how is it possible that it can edit our code at all?<\/p>\n

    This is where the harness comes in. It gives the model the power of tool calling<\/strong>, and suddenly that autocomplete can do real work: list the current directory, search the whole project for inject(<\/code>, open a file, apply an edit, run ng build<\/code>. Instead of only writing text back to us, the model emits a structured tool call, the harness actually runs it, and then feeds the result back into the conversation. (Simon Willison has a really nice, no-nonsense walk-through of how coding agents work<\/a> if you want to go one level deeper.) So in short: the harness is the wrapper around the model that lets it read our projects and then actually change them.<\/p>\n

    And that feedback loop is the whole trick. A coding agent is really just our LLM running in a loop<\/strong>: it proposes an action, the harness runs the tool and hands the output back, the model reads that output and decides what to do next \u2013 open another file, fix the error it just caused, run the tests again. This loop is what most people mean today when they talk about AI agents<\/a>: the harness manages the context, the tools, the prompts, and the control flow, while the model does the actual reasoning.<\/p>\n

    My personal take? The harness, not the model, is where most of the day-to-day quality comes from. Two developers can run the exact same Opus or GPT model and still get wildly different results, simply because one harness has the right tools, the right project context, and sensible guardrails \u2013 and the other doesn't. And that's exactly why this post exists: the harness is the part you<\/em> control, so setting it up properly for Angular<\/em> is not some nice-to-have. In my opinion, it's the highest-leverage thing you can do.<\/p>\n

    Model, Harness, App \u2013 What's the Difference?<\/h2>\n

    We now know what a harness does \u2013 but it's easy to confuse it with the model underneath it and the app around it. So let me make that concrete with the exact setup I've been using for Angular<\/em> development today:<\/p>\n