With everyone going on about agentic AI and automation, I thought it might be good to take a look at a tool that creates these linked-together actions that isn’t super-technical and complex. Glif is basically agentic AI for beginners and I really like it.

You get 10 free credits per day, but they don’t roll over which is a bit of a pain when the glif you’d like to run takes 11 credits or so. What is useful is that this isn’t a subscription tool. You buy bundles of credits and the smallest one is 200 credits for US$1.99. I used this bundle to build my example glif, run it a bunch of times and I still have 94-odd credits left.

You can build a glif for free, you just can’t run it without credits… And you’ll see in the components of my glif below, the interface gives you credit estimates for each component as you build, so you know what you can ‘afford’.

I created a simple image generator, imaginatively called Pretty Image Generator. It takes a text prompt in, then pulls that into ChatGPT to expand the prompt and add some consistent elements, then puts it into an image generator. Pretty simple, and you could do all of it in Chat GPT, but glif has a couple of advantages:

1. You have an insane list of LLM models to pick from, all the models from the paid levels of the big ones, plus more, and the same for the image generators (they have varying credit costs, so if you can build a flow for under 10 credits, you won’t need to pay).
2. It allows you to add elements to the final prompt that means your output shares characteristics (in my case that the images are beautiful and detailed, and use a warm, soft colour palette). Useful for branding…

You can see it’s a simple flow diagram, with options and controls you can play with in each section. This is what makes it more user-friendly in my view. I can see what’s going on, to the point where, once I run the glif, I can see the expanded prompt at the bottom of the second box, and then the preview image at the bottom of the last one.

Example run

I recently read a very cute, gentle book called The Full Moon Café. It’s Japanese and involves a magical pop-up café staffed by cats, so I thought I’d see what that inspired…

Input Text: a pop-up café operated out of a food truck in Kyoto. The café is run by cats, and only appears on nights with a full moon

The ‘permanent’ prompts I have in the second/LLM box:

• ChatGPT 4o system prompt – You are a talented and highly experienced graphic artist.
• Main prompt – Improve the following image prompt by expanding on the prompt to generate a beautiful, detailed image with a preference towards soft, shades of peach, blush, and gold. Prompt: “{input1}”

Resulting enhanced image prompt (and yes, it got cut off): Create a whimsical, enchanting scene set in the heart of Kyoto during a night with a luminous full moon hanging in the sky. The image captures a pop-up cafe operated out of a charming, vintage-style food truck, artfully decorated with soft shades of peach, blush, and hints of gold. The truck is parked under a canopy of cherry blossom trees, their delicate pink petals softly illuminated by the moonlight, blending harmoniously with the cafe’s color palette. In this magical setting, the food truck is bustling with activity, run entirely by an ensemble of adorable cats. These feline baristas and chefs are depicted with delightful, expressive features, each wearing tiny aprons and hats that match the cafe’s aesthetic. Some cats are seen expertly preparing drinks and pastries, while others gracefully serve a small group of amazed patrons seated at elegant, makeshift tables decorated with dainty lanterns. The image should emphasize the serene, dreamlike atmosphere, with the soft glow of the moonlight casting gentle reflections on the

What I’d do differently: Put a word or character limit into the main LLM prompt so things don’t get missed. This can get important where you’re looking at branding and/or consistent elements across images. If you want to include a cat in a the background of all your images, you want to make sure that gets into the final image prompt, and isn’t cut off by the image generator’s character limit.

To give you an idea of how the main prompt responded to some very different inputs, take a look at the images below…

Glif can obviously run a range of different automations, around image generation, research, text output and so on. If you want to get your head around agentic AI, this could be a nice place to start.