Agentic Jackal: Live Execution and Semantic Value Grounding for Text-to-JQL

TL;DR

This paper introduces Jackal, a large-scale benchmark for translating natural language to Jira Query Language (JQL), and proposes Agentic Jackal, an agent that improves translation accuracy by executing queries live and resolving ambiguities.

Contribution

The paper presents the first execution-based benchmark for NL to JQL translation and introduces an agent that significantly enhances LLM performance through live query execution and semantic retrieval.

Findings

Single-pass LLMs average 43.4% accuracy on JQL translation.

Agentic approach improves 7 of 9 models, with a 9.0% relative gain.

Categorical-value accuracy increases from 48.7% to 71.7% with the agent.

Abstract

Translating natural language into Jira Query Language (JQL) requires resolving ambiguous field references, instance-specific categorical values, and complex Boolean predicates. Single-pass LLMs cannot discover which categorical values (e.g., component names or fix versions) actually exist in a given Jira instance, nor can they verify generated queries against a live data source, limiting accuracy on paraphrased or ambiguous requests. No open, execution-based benchmark exists for mapping natural language to JQL. We introduce Jackal, the first large-scale, execution-based text-to-JQL benchmark comprising 100,000 validated NL-JQL pairs on a live Jira instance with over 200,000 issues. To establish baselines on Jackal, we propose Agentic Jackal, a tool-augmented agent that equips LLMs with live query execution via the Jira MCP server and JiraAnchor, a semantic retrieval tool that resolves natural-language mentions of categorical values through embedding-based similarity search. Among 9 frontier LLMs evaluated, single-pass models average only 43.4% execution accuracy on short natural-language queries, highlighting that text-to-JQL remains an open challenge. The agentic approach improves 7 of 9 models, with a 9.0% relative gain on the most linguistically challenging variant; in a controlled ablation isolating JiraAnchor, categorical-value accuracy rises from 48.7% to 71.7%, with component-field accuracy jumping from 16.9% to 66.2%. Our analysis identifies inherent semantic ambiguities, such as issue-type disambiguation and text-field selection, as the dominant failure modes rather than value-resolution errors, pointing to concrete directions for future work. We publicly release the benchmark, all agent transcripts, and evaluation code to support reproducibility.