How to give LLM access to per-module in and output

Questions
1

:bullseye: What is your goal?

I keep falling back to the same issue where i cant give claude code access to per module in-and output information of completed scenario runs, even though it would be the absolute fastest way for it to debug and improve my scenarios. So now any time something fails i ask it which modules specific output it would like to see to be able to provide a solution.
The only thing it seems to have access to is generic success/failure info but not per module info.

There HAS to be some cli/api/mcp path which reveales this information but i cant find it. Anyone got a solution?

:thinking: What is the problem & what have you tried?

xecutions_list / executions_get / executions_get-detail Dont list this info

And the other path, GET /dlqs/{dlqId}/bundle only reveals this info in failed runs.

I need it from succeeding runs as well.

3

Hi @Coldsilver go check out Make MCP Server | MCP Server | Make Developer Hub

Make has MCP / API and CLI available so your clode code / codex can help debug and build. You can also use Make as a connector or via MCP in ChatGPT or Claude WebUI

4

Hi Neill,

Thanks for the reply! the MCP and CLI are nice, but they only reveal scenario level in and output. They don’t go as deep as per-module individual in and output. If I’m wrong correct me but the MCP and cli are the first places I looked

5

You are correct. Even with the API, mostly Metadata coming back of overall scenario logs. @Valery.Mezencev any inputs?

@Coldsilver do you have access to (MAIA) aka Make Co Worker in your account? The Co Worker can help with this in the UI, but you have a really good question and a valuable feature, I would also like to use with Codex / Claude Code.

image
image504Ă—726 29.3 KB

6

What is MAIA and where do I find her? I’ve never seen that interface from your screenshot. My workaround for now for claude to be able to work with per module output is data stores in the scenario catching all information. It would still be nice to have it integrated into the mcp/cli though!

7

MAIA is still in closed beta, this window will be available directly in the canvas once its released.

8

Make intentionally does this, it is not simply a gap in the API docs. It only retains per-module bundle data for failed runs because that is where debugging requires it. When the run is successful that granularity gets discarded once execution completes. The DLQ endpoint exposing it on failures is the exception and not the rule.

You’ve already hit the ceiling of what’s exposed, things like “executions_get_detail”, “executions_list” are not surfaced for successful runs.

The best workaround for this is to instrument the scenario itself. Add a data store write or webhook at key modules capturing the output while the run is live. It will take more setup upfront, but gives you exactly the per-module visibility you’re after and Claude actually works with it.

It is definitely worth submitting a feature request as this is something that comes up fairly often.

9

Hi y’all,

first of all, thanks for the valuable feedback. Happy to see that you’re leveraging these advanced features of the Make MCP Server. You’re right that module output data isn’t currently exposed through any of our tools.

That said, as we continue expanding the capabilities of our MCP Server and developing more tools for Make AI Agents, we’re exploring additional tools for Scenario Debugging to provide deeper visibility into execution data and help Agents troubleshoot more effectively. Stay tuned :rocket: