Archon: Autonomous Architecture Sentinel for Software Teams

Empowering software teams to ship with confidence: Archon audits Jira & Bitbucket code against Confluence ADRs, stopping drift with 1-click remediation, Rovo Agent skills & Atlassian MCP.

Comment

đź’ˇ Inspiration: The Velocity-Governance Gap

Every software engineer knows the feeling: it is two days before the end of the sprint, the backlog is packed, and leadership needs a critical microservice shipped yesterday. With AI coding assistants, we can now write hundreds of lines of code in seconds. We feel faster and more productive than ever—until something silently breaks in staging, or worse, triggers a late-night security alert in production.

The problem isn’t that developers don't care about security or clean design. The problem is that crucial architectural decisions live in places developers rarely look while writing code.

Principal architects spend weeks drafting thoughtful Architectural Decision Records (ADRs) in Confluence—mandating protocols like Mutual TLS (mTLS), strict data isolation, or forbidden ports. But when a developer is heads-down inside their IDE or opening a pull request, they simply do not have the time to search through 50-page wiki spaces to check if their transport client complies with policy.

As teams sprint ahead, code silently drifts away from architectural intent. By the time that drift is uncovered—usually during a high-friction security review or after an outage—fixing it requires massive refactoring, delays releases, and burns out engineers.

We built Archon Sentinel because we believe developers shouldn't have to choose between shipping fast and building securely. Archon acts as an invisible, supportive guardrail that bridges documentation and code, embodying the Codegeist 2026 mission: "Build with context. Ship with confidence."

🛡️ What It Does

Archon continuously connects the dots across Jira, Confluence, and Bitbucket so developers never have to guess whether their code meets architectural standards:

  1. In-Context Pre-Flight Gate (Jira Custom UI): When an engineer opens a Jira work item (like KAN-1), Archon's issue panel inspects the ticket and linked repositories against governing Confluence ADRs (such as ADR-042: Mandatory Mutual TLS for Inter-Service Communication).

  2. Real-Time Code Drift Detection: Archon evaluates pull requests and code commits in Bitbucket. If someone accidentally commits unencrypted cleartext transport (like plain http on port 8085), Archon instantly flags a clear, human-readable VIOLATION DETECTED alert right on the ticket.

  3. Strategic Technical Debt Telemetry: It translates technical drift into business reality, showing estimated remediation debt against the team's cloud security budget so engineering leads have instant visibility.

  4. 1-Click Interactive Auto-Remediation: Instead of blocking developers or handing them a wall of text, Archon provides an actionable ⚡ Apply Auto-Remediation (mTLS) button directly inside Jira. One click updates the transport configuration to an encrypted mTLS context, immediately turning the status badge to a clean green COMPLIANT.

  5. Assignable Rovo Agent & MCP Tooling: Archon is a native Rovo Agent (archon-agent) with custom Rovo Skills (architectural-compliance-skill). Developers can ask questions directly in Rovo Chat or summon Archon inside external AI IDEs (like Cursor or Claude Code) via our Model Context Protocol (MCP) server before committing code.

⚙️ How I Built It

Archon was built 100% serverless on the Atlassian Forge platform (nodejs22.x), ensuring enterprise data sovereignty and zero infrastructure overhead:

Atlassian Forge & Custom UI: Built using @forge/bridge and @forge/resolver with a responsive React frontend that gives developers a clean, interactive issue panel with real-time state transitions.

Confluence Cloud Integration: Securely queries Confluence REST APIs using authenticated asUser() context to evaluate active Architectural Decision Records under the user's authentic permissions.

Bitbucket Source Integration: Connects pull request branches (archon-violation-test) and code diffs directly to Jira issue keys.

Native Assignable Rovo Agent (rovo:agent):Declared with specialized prompts and structured instructions defined in SKILL.md frontmatter.

Atlassian Model Context Protocol (rovo:mcp): Configured with dedicated action modules, exposing verify-architectural-alignment as a remote MCP tool for external AI agents.

Runs on Atlassian Security:100% Forge Hosted Storage with zero external network egress—no proprietary company code, tickets, or user data ever leave Atlassian Cloud.

🧗 Challenges I Faced

  1. Marketplace Security Scans & User-Level Authorization: During our initial production deployment for Marketplace review, Atlassian's automated scanner (FSRT / EcoScanner) flagged an authorization check (AMS-72250). Resolving this taught us a great deal about enterprise Forge security: we eliminated elevated asApp() calls, transitioned our resolvers strictly to authenticated asUser() context, and pruned unused write scopes to follow strict least-privilege standards.

  2. Connecting Context Across Disparate Products: Ensuring that a code diff in Bitbucket, an issue in Jira, and a published ADR in Confluence stayed synchronized in real time without lagging the UI required careful caching and clean fallback mechanisms.

  3. Mastering the Rovo & MCP Manifest Schemas: Aligning rovo:agent, rovo:skill, and rovo:mcp within Forge's strict property schemas, character limits, and YAML frontmatter conventions required meticulous testing and iterative deployments.

🧠 What I Learned

Context Turns AI into an Asset, Not a Liability: AI models can generate thousands of lines of syntax, but they don't know your company's rules, security requirements, or architectural history. Giving AI tools access to connected organizational context (like Confluence ADRs) is what makes autonomous engineering practical.

The Speed and Power of Forge: Building on Atlassian Forge let us go from an idea to a multi-product, secure, serverless application with built-in OAuth and zero server management.

🚀 What's Next for Archon

Deeper Teamwork Graph Traversal: Moving from REST endpoints to full Cypher graph queries as Atlassian's Teamwork Graph capabilities expand.

Expanding Policy Types:** Broadening our automated code evaluation beyond transport protocols to include Kubernetes manifests, Terraform infrastructure-as-code files, and database migration scripts.

Public Marketplace Launch: Following our initial production submission (ECOHELP-180595), we look forward to completing partner onboarding and offering Archon to software engineering teams worldwide.

What we learned

What's next for Archon: Autonomous Architecture Sentinel for Software Teams

Built With

Share this project:

Updates

Submission history