All-in-One Pickaxe (Multi-Model Setup)

[abhi]

Over the last few weeks, while chatting with many community members, I kept hearing the same question:

Can a single Pickaxe work as an all-in-one chatbot and switch between different AI models based on the user’s query?

Short answer: yes. And it works really well.

I built this setup, have been using it actively, and it’s been running smoothly without any issues. No code, no hacks. Just Pickaxe Actions used the right way.

I’m sharing the exact setup here so anyone in the community can build the same thing in minutes.

What this setup does

• Users chat with one main Pickaxe

• They mention which model they want (GPT-5, Gemini 3 Pro, etc.)

• The main Pickaxe triggers the correct Action

• The answer comes back from that model automatically

Step by step setup

Step 1: Create the main Pickaxe

• Create a chatbot Pickaxe

• Name it All in One (or any other preferable name)

• This is the only Pickaxe users interact with

Step 2: Create a worker Pickaxe

• Create a second chatbot Pickaxe

• Set its model to Grmini 3 Pro

• Keep the prompt simple. Just answer the question

  

Step 3: Clone for other models

• Clone the Pickaxe. Go to the Studio, click on “Duplicate to” and clone it within the same Studio. Rename cloned duplicate pickaxes for easy identification

• Change the model for each clone

  • One uses Gemini 3 Pro

  • One uses GPT-5

  • Add more models if you want

Each Pickaxe equals one model

Step 4: Connect everything with Actions

• Open All in One: Pickaxe Chatbot

• Go to the Actions tab

• Add one Action per worker Pickaxe

For each Action, define a clear trigger prompt

Example

• “If the user says ‘use Gemini 3 Pro’, run this Action and send the question to the Gemini 3 Pro Pickaxe”

• “Call this action if user requests output using Gemini 3 Pro or Gemini3 Pro or Google Gemini 3 Pro.”

• “If the user says ‘use GPT-5’, run this Action and return the answer”

Clarity here matters. Use exact model names and variations to be on the safe side.

How it works in practice

User types:

• “Use Gemini 3 Pro and answer 2+2”

Result:

• The Gemini 3 Pro Action triggers

• The worker Pickaxe runs

• The answer is returned in the main chat

Same flow for any other model.

How to confirm it’s working

• Go to your Studio

• Open Message analytics / Message insights (follow screenshot steps)

• You can see whether the Action was triggered

This helps you quickly confirm that the Action ran, and shows the actual cost so you can price your Pickaxe accurately.

Reliability & Cost:

I’ve been running this setup for the past few weeks without any issues.
It’s already been used by 150+ users across 1,500+ runs, with a total cost of roughly $25.

Sharing this in case it helps others build something similar or answers the all-in-one Pickaxe question once and for all.

[hurmuli]

Are all models connected to each other and what is the benefit of this instead of just connecting the other pickaxes to one main Pickaxe?

[abhi]

@hurmuli

Are all models connected to each other and what is the benefit of this instead of just connecting the other pickaxes to one main Pickaxe?

Totally fair question.

The models don’t need to be connected to each other. In this setup, each model sits in its own Pickaxe and the All in One Pickaxe acts as the front door.

You can connect models to each other if you want, but for most cases it’s not really necessary.

The main advantage of using one All in One Pickaxe is flexibility and a cleaner user experience. For example, I have users who prefer:

• Claude for writing tasks

• Grok for social media posts

• ChatGPT for general questions

I could give them three separate Pickaxes, or I can just give them one All in One Pickaxe and let them choose the model in the prompt.

From a builder point of view, this also helps with pricing. You can offer a premium all-in-one Pickaxe, or keep simpler single-model Pickaxes cheaper. Same building blocks, different packaging based on what your users want.

[Ned.Malki]

@abhi Nice! This can also be used as an “LLM council” - similar to the most recent feature Perplexity implemented.

[kenlyle]

This approaches the auto-router functionality I was imagining.
If we could add something like LLM Leaderboard - Comparison of over 100 AI models from OpenAI, Google, DeepSeek & others to the KB of the All-in-one PA, with the right prompt, might we be able to get that PA to do auto-routing?
Can any of the experts help push this forward?

[user14]

So what model did you use for the ‘Main’ pickaxe? If the user doesn’t specify a model, I presume then this is the fallback? Do you have a ‘lead in’ that actively asks what model they want to use?

[abhi]

I use GPT 5 for the main router. Yes, that also becomes a fallback model. You can customize the intro message to make sure people understand how to use it.

[PROVIDER_DAUWZ00NXM]

Hi Abhi,

Is it possible for the All-in-One pickaxe to be a form + chat interface where you could have options like Gemini, ChatGPT, Claude, etc.?

And, using a specific prompt, the user could choose which model to use?

Thanks in advance,

Vasco

[abhi]

Hi @PROVIDER_DAUWZ00NXM, it should work.