Just released: Ash AI, a new extension for Ash Framework. LLM integration for Elixir apps. Use Ash’s declarative approach for structured outputs, secure tool calling & MCP.
Read the blog post for more! Ash AI: A comprehensive LLM toolbox for Ash Framework
37 Likes
a development tool MCP server which complements Tidewave perfectly
Could you outline an example for this? 
Haven’t yet tried using them in tandem and I’m wondering how would I do that.
AFAIS Tidewave acts as a general MCP server while Ash AI goes through Ash.
I imagine if I want MCP to fetch something Tidewave would access the database directly while Ash AI would go through a read action - is this correct?
1 Like
There are three separate MCPs in this conversation:
- tidewave - this is a dev MCP that can do pretty much “anything” within your running application (like read from the database)
- the MCP server tooling we provide that lets you expose your actions as tools over MCP for external consumption
- the dev tools MCP server, which just currently has 2 experimental tools (
list_resources
and list_generators
)
3 Likes
Hi @zachdaniel ,
thanks for looking into adding this great addition to Ash as it is one of the things I believe will be in demand in the future “Making software ready for AI agents customers”.
I have tried the chat one as a sample and I just would like to confirm with you what it can provide, I tried it with just a sample new project and I noticed it gives me just general answers right now, is this the intent? Or the intent to have a new interface to services based on prompt? as an example “prompt : generate an invoice to the customer ABC and send it to the e-mail”?
One amazing thing, I switched to another language and right away it answered

below image shows what I got now.
You can add tools by modifying the response change. See the readme for how to make your ash actions callable as tools
then, alter the code in your response change like tools: [:your, :tool, :names]
. Then your agent can use your app via those tools 