If I ask say Claude to run a report or create a chart of the data in say an airtable, it will create something nice.
If I do the same in Pickaxe, you just get text back.
Is there a way or an action that can make the output better ?
If I ask say Claude to run a report or create a chart of the data in say an airtable, it will create something nice.
If I do the same in Pickaxe, you just get text back.
Is there a way or an action that can make the output better ?
Hi,
Great question!
When you use Claude on their own platform, they have hidden system prompts and formatting instructions running behind the scenes that help produce more polished outputs like nicely formatted reports and charts. Unfortunately, these hidden prompts are not included when you access the model through API keys or third-party tools like Pickaxe.
However, you can still achieve similar results in Pickaxe. You can instruct your bot to generate responses in Markdown or HTML so you have more control over formatting and layout. You can also use the Chart Generator to create visual charts directly from your data. Additionally, the PDF Generator can help you build downloadable reports in a more structured format.
Hey @ihmunro, long time. Hope you’re well!
Great question. What you’re noticing is a key difference in how most LLM-based tools (like ChatGPT/Claude) handle output versus how Pickaxe works when connected to external data like Airtable.
By default, Pickaxe returns text because the core LLMs themselves are text-in, text-out systems. However, Pickaxe is designed to go beyond this with Actions. Here’s the strategic angle:
Next steps: