Inspiration
In modern software engineering, developers write code rapidly inside AI-assisted IDEs (Cursor, VS Code, Claude Code), while product specifications, acceptance criteria, and architecture decisions remain siloed in Jira tickets and Confluence PRDs. This disconnect leads to "spec drift"—features get implemented, but edge cases and compliance constraints are missed until QA or production. We built CodeContext SpecGuard to bridge developer IDEs directly with Atlassian context.
What it does
CodeContext SpecGuard creates an automated verification and traceability loop for software teams:
- Model Context Protocol Server (
rovo:mcp): Exposes live Jira issue specifications and Confluence PRDs directly into developer IDEs via MCP. - Autonomous Rovo Agent (
rovo:agent): An intelligent QA sentinel inside Jira that verifies code diffs against acceptance criteria, flags missing requirements, and calculates a compliance score. - Native Jira Issue Panel (
jira:issuePanel&jira:issueContext): Embedded UI Kit 2 dashboard inside Jira issue views that automatically extracts criteria from ADF descriptions, runs instant audits on submitted code, and records audit reports as official Jira comments.
How we built it
- Platform: Atlassian Forge runtime (Node 24, ARM64).
- Frontend: Atlassian UI Kit 2 (
@forge/react) using native reconciler and@forge/bridge. - Rovo & MCP: Integrated
rovo:mcpexposing 5 unified tools androvo:agentwith custom system prompts for requirement traceability. - Backend: JavaScript with custom Atlassian Document Format (ADF) parser and heuristic keyword compliance engine.
- Testing: Native Node.js test runner covering ADF extraction and diff auditing.
Challenges we ran into
Parsing deeply nested Atlassian Document Format (ADF) structures into discrete, testable acceptance criteria required building a robust recursive text-normalization pipeline. Additionally, ensuring seamless interop between modern ESM imports and Forge runtime resolvers required careful architectural structuring.
Accomplishments that we're proud of
- Seamless dual-surface architecture: the exact same 5 backend tools serve both developer IDEs via MCP and the autonomous Rovo agent in the cloud.
- Instant, non-intrusive UI Kit 2 panel inside Jira with real-time visual progress bars and zero third-party tracking dependencies.
What's next for CodeContext SpecGuard
- Bi-directional sync with Bitbucket automated merge checks to block pull requests whose compliance score falls below 80%.
- Deeper integration with Atlassian Teamwork Graph (TWG) to track cross-project requirement drift over time.
Built With
- atlassian-forge
- confluence
- javascript
- jira
- mcp
- node.js
- react
- rovo
- typescript