Title: Best way to connect a Pickaxe chat + form to a client’s proprietary calculation tool (send inputs -> get graphs back -> continue chatting)

Hey Pickaxe community - looking for guidance on an integration pattern.

Scenario: My client has a proprietary tax calculation tool. Today it’s purely form-based: user inputs numbers → tool crunches inputs → outputs charts/visualizations. Historically it was static (no chat, just results).

We built a Pickaxe experience where users can ask questions about their taxes and get guidance in chat.

What we want next: We want Pickaxe to “talk to” the client’s tool so a user can

  1. Chat with the AI or fill a Pickaxe form with their tax numbers
  2. Those numbers get sent to the client’s tax calculation tool
  3. The tool returns calculation results + graphs
  4. Those graphs get shown inside the Pickaxe chat experience
  5. The user continues chatting, referencing those outputs (e.g., “Why did this spike?”, “How would scenario B change it?”)

Question: What’s the best way to implement this with Pickaxe?

What I’m trying to understand

  1. Does Pickaxe support calling an external API (webhook / function call / integration) from either a form submission or chat?
  2. What’s the recommended architecture to:
    1. send structured inputs to an external tool
    2. receive results + chart assets,
    3. display them in the chat
    4. and store the returned results as context so the user can keep asking follow-ups?
    5. If the external tool can return either raw data or pre-rendered chart images, which approach tends to work best in Pickaxe?

Current constraints:
-
The client tool is proprietary but can expose an API endpoint if needed

  • We can implement a middleware layer if required (auth, validation, formatting, caching)

Any examples, recommended patterns, or “here’s how others do it” would be massively appreciated.

2 Likes

There is a built in Action for Generic Webhook.

Hi @lumora Yes, Pickaxe supports custom actions (function call) for external APIs. There are options for easy-to-configure Webhooks and MCP servers. You can also create custom actions and include your APIs if you Understand Python/JSON code.

Answers to Q1+2:

  1. Webhook (for single API) / MCP (for multiple APIs).
  2. A webhook/MCP can both send and recieve data.
  3. That will happen automatically as long as your actions are configured correctly.
  4. This automatically happens in the chat as long as you set the memory for the bot to be long enough (taking into consideration, what the average length of the chats would be). Data will be stored in chat history once user closes that chat (As long as you have chat history enabled).
  5. Can you clarify what you mean by raw data here? In terms of generating chart images, this is possible. Here’s a video I made on this:

Hope that helps!

-Ned