Monkey SQL

Monkey SQL extends Rovo’s JQL-style verification to all Jira data, not just issues.

Comment

Inspiration

We loved how Rovo solved a fundamental reliability challenge: exposing the JQL behind each answer shows exactly what the AI understood and lets users verify results in the Issue Navigator. But JQL only covers issues, so we wanted the same transparency across all Jira data.

What it does

Monkey brings the same clarity that Rovo provides with JQL to all of Jira, not only issues.

How we built it

  1. Monkey transforms Jira into a virtual database with support for standard SQL to query any data (not only issues).
  2. We then trained an AI model (ChatGPT 4.1-mini) on thousands of prompts and SQL queries to turn natural language into precise, auditable SQL queries.
  3. Rovo ties it all together as the interactive AI interface, enriching user intent with Jira context and turning questions into fast, transparent answers.

Workflow:

User's prompt in NL (Rovo) -> Fine-tuned AI model (OpenAI) -> Virtual SQL query (Forge) -> Jira REST API -> Jira data

Challenges we ran into

Balancing a clean Rovo interaction with the messy reality of Jira’s REST APIs and data structures.

Accomplishments that we're proud of

We delivered an AI agent that enables Rovo users to obtain verifiable, auditable answers across Jira data, with full transparency into how results are produced. By exposing Jira’s data model through a structured, queryable layer, we help extend Rovo’s search and question-answering capabilities while preserving trust, traceability, and enterprise-grade security.

What we learned

The most significant takeaway was proving that AI fine-tuning doesn't have to be expensive or slow. We learned that training against a defined data model abstraction (SQL schema) is far more efficient than using large, curated datasets, and at a fraction of the cost.

What's next for Monkey SQL

We are re-engineering our SQL engine to replace the REST API with high-performance, highly efficient GraphQL. This will cover the entire Atlassian suite (Jira, Confluence, Bitbucket, etc.) and any data sources connected to TeamWork Graph.

Built With

Share this project:

Updates