Agent Tools

The Jira agent provides 4 tools for searching and analyzing tickets:

Tool	Description	Key Parameters
deep_search_jira	Multi-strategy search that generates 3-4 JQL variations from natural language, executes them in parallel, and deduplicates results.	query, project, max_results
get_overdue_issues	List issues past their due date, sorted by priority (highest first) and due date (oldest first).	project, priority_filter, max_results
summarize_recent_activity	Fetch recently updated issues and generate an LLM-powered summary of activity.	project, time_window
get_issue_details	Retrieve full context for a specific issue: description, status, assignee, comments, and change history.	issue_key, include_comments, include_changelog

Query Enhancement

The deep_search_jira tool uses an LLM-powered query enhancer that transforms natural language into multiple JQL variations:

1. Analyze — The query enhancer extracts keywords and identifies synonyms from the user’s query
2. Generate — It creates 3-4 different JQL queries with varied text search fields, status filters, and project scopes
3. Execute — All variations are executed against Jira
4. Merge — Results are deduplicated by issue key

If the LLM fails to generate valid JQL, the system falls back to basic text search variations automatically.

Manual Override

Prefix your query with /jira to route directly to the Jira agent, bypassing the workflow classification:

/jira What's the status of the authentication refactor?

When the Jira agent fails to find relevant results, it automatically suggests using the RAG agent as a follow-up.