Context

[mdlmarkham]

This tool works very well - especially for creating or updating a single ticket. I'm finding that I often want to have the Agent review a backlog of many tickets - looking for opportunities to consolidate, close things out, re-assign them etc.

When I do this I often have context window issues with Claude Desktop... presumably because the text of many tickets is being returned. What could be done to manage this?

[Troubladore]

A few things that can help with context window usage when working with large backlogs:

1. Use the fields parameter on jira_search to request only what you need — e.g. fields: "summary,status,assignee,priority" instead of the default which includes descriptions. This significantly reduces the data returned per issue.

2. Use limit and pagination — the search tool defaults to 10 results at a time (max 50). You can page through results with start_at rather than pulling everything at once.

3. Use the TOOLSETS environment variable to expose only the toolsets your workflow needs (e.g. TOOLSETS=jira_issues,jira_search), which reduces the tool definitions in the context.

The combination of selective fields + smaller pages should keep individual responses compact enough for the agent to work through a large backlog iteratively.