Re: Exploring MCP-based tools for Incubator insights

Hi Justin

That's pretty cool! Thanks!

What do you think also about adding a CLI on top of the MCP. The reason I
think CLI is interesting is because it's easier to integrate with Claude
Code :)
Happy to work on something (as I have a prototype in progress in Apache
Karaf).

Regards
JB

On Sun, Apr 19, 2026 at 3:55 AM Justin Mclean <[email protected]>
wrote:

> Hi,
>
> I have been experimenting with a small set of tools to explore how we
> might better understand Incubator data and podling signals in a more
> queryable way.
>
> The approach is based on simple MCP (Model Context Protocol) servers that
> expose different layers of Incubator information:
> - PodlingsMCP: lifecycle and metadata from podlings.xml
> - HealthMCP: parsed metrics from the incubator health reports
> - IncubatorMCP: combines the above to produce IPMC-oriented views such as
> watchlists, readiness signals, and podling briefs
>
> There is also a small skills file included to make it easier for MCP
> clients to use these tools.
>
> The intent is not to replace existing processes or judgment, but to make
> it easier to:
> -  identify podlings that may need attention
> - spot patterns across reports and activity
> - generate quick summaries for discussion
> - experiment with what the data suggests before applying human context
>
> The higher-level layer (IncubatorMCP) keeps a clear separation between:
> -  source facts (from podlings.xml and health reports)
> - derived signals (heuristics around activity, mentoring, reporting, and
> so on)
>
> This is still early and intentionally lightweight, but it may be useful as:
> - a way to explore Incubator data programmatically
> - a potential aid for reporting or mentoring workflows
> - a foundation for further experimentation
>
> Repositories are here if anyone is interested:
> -  https://github.com/justinmclean/PodlingsMCP
> - https://github.com/justinmclean/HealthMCP
> - https://github.com/justinmclean/IncubatorMCP
>
> I would be interested in feedback on a few things:
> - whether this kind of tooling would be useful for IPMC work
> - whether the signals approach aligns with how people think about podling
> health
> - any obvious gaps or misinterpretations in the heuristics
>
> Note that it can also produce many dashboard-style displays and graphs of
> the data it has access to.
>
> Thanks,
> Justin
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
>
>