Skip to content

Conversation

@jonathanhefner
Copy link
Member

@jonathanhefner jonathanhefner commented Nov 14, 2025

Follow-up to #1732.

This document explains how tasks can be used in some motivating use cases.

I'm not sure of the best place for this document, or whether it should be merged into another document. Opening this PR as draft to collect feedback.

(To view the rendered Mermaid diagram, see https://github.com/modelcontextprotocol/modelcontextprotocol/blob/9a8401d3938c5dcc7a674cb217f7ab30a50ada2f/docs/tasks-uses-cases.mdx.)

@jonathanhefner jonathanhefner mentioned this pull request Nov 14, 2025
9 tasks
2. Delivers the rejection as a system message to the LLM
3. The LLM adapts by using the built-in `fetch` API instead

## Why This Maps Well to Tasks
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this section's content could maybe be merged into "Key Takeaway"? Something along the lines of, "tasks map naturally to this flow for these reasons... this same pattern applies across multiple domains".

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I actually like having these two as separate sections. I think they serve different purposes: one explains why tasks are a good tool for the use case, the other identifies similar use cases.

I'll give it some more thought though.


This same pattern applies across multiple domains:

**Data analysis**: Submit computational jobs (molecular analysis, drug interaction modeling) that process for hours while the LLM continues other work.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I would probably use other examples in this parenthetical, like maybe database migrations and benchmarks? I know I used these as examples in the motivation section of #1686, but these are oddly specific in documentation.

I know Claude really loves pulling out these particular use cases from that issue, though 😅

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think molecular analysis and drug interaction modeling are good examples for the category of "Data analysis", and they provide some variety compared to the other categories. But I've inserted an "e.g." to clarify that they are just examples.

@LucaButBoring
Copy link
Contributor

LucaButBoring commented Nov 15, 2025

Left comments, this looks good overall, I just had some feedback on a few minor phrasing points and the diagram.

@jonathanhefner jonathanhefner force-pushed the tasks-use-cases branch 2 times, most recently from 73c3103 to 034ea6f Compare November 15, 2025 03:06
Follow-up to modelcontextprotocol#1732.

This explains how tasks can be used in some motivating use cases.

Co-Authored-By: Claude <noreply@anthropic.com>
deactivate LLM
Client->>LLM: Files updated
activate LLM
```
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Client begins polling task statuses need some interval, the current shape seems always polling in a while loop

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants