How to Leverage AI for Smarter ProcessWire Development

[bernhard]

On 2/12/2026 at 10:52 PM, Peter Knight said:

Don’t forget Cursor cloud agents.

Give a cloud agent a plan and the agent works in the background while you’re away from the laptop.

I give them a task on the iPad when I’m at the gym and by the time I’m home there’s a new branch waiting for me. 

I can’t verify this but somehow I’ve often found the quality of the cloud agents to be better than the desktop ones even on the same model. Can’t be true? But feels that way.

Oh actually I completely missed that! Would be interested in more details how you do that or any helpful links 🙂 

[Peter Knight]

Sure, just visit cursor.com on a mobile device and log in.

Look for the Agents link in top right. 

Select your project and branch if you’re connected to Git

Then if you’re out, at the gym or on anywhere away from desk I just dictate the features I want and let the cloud agents run.

[Peter Knight]

Anyone tried OpenClaw yet on a local machine? I see lots of tweets about Mac minis and OC and people running much cheaper agents etc 

[elabx]

10 hours ago, bernhard said:

Oh actually I completely missed that! Would be interested in more details how you do that or any helpful links 🙂 

I though you were talking about THIS, now  I just furiously started prompting all the bugs meanwhile I go reviewing and doing stuff on my own.

Now just boostrapping ddev somwhere (gihub codespaces?) , some E2E testing, feedback loop till acceptance and out of work lol

[bernhard]

Thx @Peter Knight ! This is so crazy!

I asked it to add an onboarding tour overnight. The first two page impressions led to exceptions, but after copying those error messages to cursor it worked:

And it not only works it also looks great! There are some issues still, but it's amazing how fast you can try things out and get a real world experience and not just a pencil sketch.

Is it somehow possible to give the AI access to a browser so that it can try to load the page and see and fix such exceptions on its own? Another thing I'd love to have is to make it PLAN upfront but then start building once that is done. It seems I have to always confirm the build step after the plan is done.

[elabx]

2 hours ago, bernhard said:

Is it somehow possible to give the AI access to a browser so that it can try to load the page and see and fix such exceptions on its own?

You mean something like this?? https://cursor.com/docs/agent/browser I started doing this with Claude and it does catch the error screens and fixes the bug. 

[elabx]

Worked on this the last couple days: https://github.com/elabx/processwire-mcp just tested it yesterday, so not a lot of usage.

[gebeer]

On 2/21/2026 at 6:29 PM, bernhard said:

Is it somehow possible to give the AI access to a browser so that it can try to load the page and see and fix such exceptions on its own?

I use https://github.com/ChromeDevTools/chrome-devtools-mcp for that. Very fast.

On 2/21/2026 at 6:29 PM, bernhard said:

Another thing I'd love to have is to make it PLAN upfront but then start building once that is done. It seems I have to always confirm the build step after the plan is done.

The thing about the plan is that it's supposed to be reviewed before it is being applied, haha. But if you trust it without at least a quick glance, ok. I think Cursor can play sounds when it needs your attention. You could use that to notify you when you have to click the button.

 

20 hours ago, elabx said:

Worked on this the last couple days: https://github.com/elabx/processwire-mcp just tested it yesterday, so not a lot of usage.

Wow. The "Full ProcessWire API access - Query, create, update, and delete pages" is the most interesting for me here.

Working right now on a single-file PW-API-docs database based on https://github.com/memvid/memvid. Has semantic vector search (local embedding model), BM25 and all that good stuff. Also supports CRUD. I fed it a good part of https://github.com/phlppschrr/processwire-api-docs/blob/main/api-docs/index.md . File currently has around 35MB. Search is blazingly fast. I implement it as portable skill, not as MCP. Needs a little more love and testing but I'll drop it soonish.  

[elabx]

On 2/22/2026 at 5:59 AM, gebeer said:

Wow. The "Full ProcessWire API access - Query, create, update, and delete pages" is the most interesting for me here.

That's a bit of the AI bragging, still missing field/template creation to put an example. I focused first on page CRUD.

Quote

Working right now on a single-file PW-API-docs database based on https://github.com/memvid/memvid. Has semantic vector search (local embedding model), BM25 and all that good stuff. Also supports CRUD. I fed it a good part of https://github.com/phlppschrr/processwire-api-docs/blob/main/api-docs/index.md . 

This looks amazing 😮 

[bernhard]

On 2/22/2026 at 12:59 PM, gebeer said:

I use https://github.com/ChromeDevTools/chrome-devtools-mcp for that. Very fast.

And you get useful results? I installed it in cursor and it was impressive to see the browser pop up, but not really useful... I even added an /auto-login route to RockDevTools to make the mcp login as superuser by default, but still it was not able to fix such a latte exception issue.

[gebeer]

11 hours ago, bernhard said:

And you get useful results? I installed it in cursor and it was impressive to see the browser pop up, but not really useful... I even added an /auto-login route to RockDevTools to make the mcp login as superuser by default, but still it was not able to fix such a latte exception issue.

Yes, very good results. It's fast and pretty token efficient. You can connect it to an already open browser wit a logged in session etc. no need for auto-login route. let your agent read the mcp instructions. mine said to start chrome with a debug flag. Let me pulled that up through my conversation-search MCP quickly. Here it is :-)
 

Launch command (detached from terminal):
  setsid chromium --remote-debugging-port=9222 &>/dev/null &

  - --remote-debugging-port=9222 — enables CDP so the MCP can connect
  - setsid — creates a new session, fully detached from the terminal's process group (plain nohup doesn't survive
  terminal close because the terminal sends SIGTERM, not just SIGHUP)
  - &>/dev/null & — suppresses output and backgrounds it

That's for Linux but should work on your Mac, too.
In that browser you open your PW project, backend and then the mcp can connect.
That's the way it's supposed to be done, I guess. 

[gebeer]

23 hours ago, elabx said:

Working right now on a single-file PW-API-docs database based on https://github.com/memvid/memvid. Has semantic vector search (local embedding model), BM25 and all that good stuff. Also supports CRUD. I fed it a good part of https://github.com/phlppschrr/processwire-api-docs/blob/main/api-docs/index.md . 

Skill is up at https://github.com/gebeer/processwire-ai-docs/tree/main/skills/processwire-memory

You need to install memvid-cli. It's all in the README and skill.

You can build your memory file with docs you need from https://github.com/phlppschrr/processwire-api-docs/blob/main/api-docs/index.md 

If you want me to share my mem file (~35MB), I can do that, too.

I haven't used it a lot yet but it seems to work quite well. Maybe needs some work on the proactive part so that agents know when to lookup stuff even if not explicitly prompted. Implementing that depends very much on the AI tools you're using. For Claude Code hooks would be a good place. Others like Cursor, I don't know.

[gebeer]

3 minutes ago, gebeer said:

Skill is up at https://github.com/gebeer/processwire-ai-docs/tree/main/skills/processwire-memory

I wanted to add: This gives you basically something like context7. But locally with your very specific knowledge and not implemented as MCP but as skill which has less overhead in the context window.

And you could modify it easily for other knowledge. Different frameworks, whatever.

[wbmnfktr]

Not that long ago I started with a completely new approach in regards to ProcessWire an AI.

I had some rules, commands, and settings for Windsurf and Cursor months back but nothing really worked as good as I hoped it would and lost interest. Even switched to Astro, NextJS, BHVR, and other JS solutions. 🙈

ProcessWire-based development went back to 80% hand-coded with some assistance on the sidelines, mostly debugging, security, and tasks that were more PHP-focussed than ProcessWire-related.

Then AGENTS.md and SKILLS came up.

Custom instructions, guardrails, links to docs and example code. Almost similar to a README.md but for AI coding tools. The main problem still was the knowledge gap in models. They just didn't know enough about ProcessWire works and how to do more complex stuff.

With SKILLS this changed.

I pulled the entire docs, some blog posts, took some of the old recipes and let the AI do its magic.

The AI repackaged the entire docs, custom instructions, module docs, PHP best practices, and everything else into skills.

I installed new and clean instances of ProcessWire and just asked the AI to build things. Yeah... didn't work as expected.

I gave the AI more tools to work with and fixed the AGENTS.md:

• RockMigrations - for creating and updating templates, fields, and pages 🥰
• AutoTemplateStubs - for details about existing templates and fields 🤯
• ProcessDatabaseBackups - can be a good idea to give your tools a database file it can look into without the need to bootstrap ProcessWire into a custom script or similar.
• Inline comments in config.php to mark things as important or noteworthy otherwise that file would be ignored
• /init - a custom action OpenCode, Claude Code, and some other tools have to initialize the whole project 💯

Now my tools have all the skills and know how to use RockMigrations to create templates, fields, and pages, can trigger database backups and look into the made changes, know how to build a custom module.

Even custom page classes or URL hooks aren't a problem anymore.

The AGENTS.md contains now critical changes in the config.php, has links to all the skills, and whatever necessary. The /init command is very capable of creating it nowadays.

I just started to test it more and more. With fresh installations, older projects and even recent projects that have tons of everything.

Whenever problems occure I let the AI update the skills or create new ones that take care of the problems it faced. Learning by mistakes.

The overall workflow

A README.md with a scope of the project, necessary templates and their fields, overall main features besides handling page rendering, like a bookmarking function for recipes or read articles, newsletter signups, automation tasks to clean up older data, and whatever the project needs.

The amount of typing is still the same but now mainly in markdown files that explain what to build (/specs), what to fix (/issues), and what we have done already (/docs).

I always start in PLAN mode.

Starting simple with the overall idea, goals, and outcome. Then the combination of tool and model is important. OpenCode and Claude Code are great at thinking and planning but they need a capable model like Opus, Kimi K2.5 Thinking or even GLM-5. They start to ask questions, give options, recommend workflows. When that's done i ask to save everything to a file in the /specs folder. From here I can either switch to BUILD mode, manually tweaks the plan file, or let other models (Gemini 3 Pro or Codex) review it and ask for suggestions, changes, fixes.

 

Github: https://github.com/webmanufaktur/pwaiworkflow/

Installation:

1. Download and extract files
2. Move entire .agents folder into the project root next to the wire/ folder
3. Look into create-symlinks.sh, run if necessary

--- 

Plans

• Windsurf
  $10/month (Legacy) with access to all Anthropic, OpenAI, Google models, and many more
• Z.AI
  Coding Plan Pro $120/year (BlackFriday Deal) with all GLM models, including GLM-5 and future releases
• Kimi
  Moderato $19/month with Kimi K2.5 (just expired)
• Minimax
  Coding Plan $20/month with M2.5 (started using it as successor to Kimi)

Tools

• OpenCode - https://opencode.ai/ 
  Similar to Claude Code, easy to configure, and even easier to extend with custom modes, agents, skills, and whatever you might need or want. Has a great planning mode and doesn't ask unnecessary questions in the middle of tasks like Claude Code did for a while just to burn more tokens.
• Kimi Code CLI (with Kimi K2.5) - https://www.kimi.com/code/en 
  Tested it last month and while it's a CLI like OpenCode/Claude Code it feels and works totally different. It doesn't have any modes but supports AGENTS.md and SKILLS. Super fast and it is super capable for quick fixes, smaller features, or heavy automations.
• Windsurf IDE - https://windsurf.com/editor 
  Like Cursor with almost identical features, a custom terminal integration, includes a browser that has full access and control which is great for debugging, UI/UX (especially with Opus 4x.) - I guess most of you have seen in the past or even tried it. Was called Codium before and I know some of you used that Codium Extension which was awesome.