Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for ChronosMCP (VeriGraph AI)

Inspiration

In enterprise software organizations, releases rarely break production because engineers write defective code. They break production because critical operational context is trapped in silos.

When a backend team deploys an API schema migration (SCRUM-1), they cannot readily foresee that a separate mobile checkout pipeline (SCRUM-2) depends on the legacy contract, or that an unread Confluence governance policy (POL-402: PCI-DSS v4.0) mandates a 14-day backward-compatibility adapter. The result is catastrophic rollback, broken customer transactions, and emergency war rooms.

We built ChronosMCP (VeriGraph AI) to turn Jira into an autonomous flight simulator for releases—combining Atlassian Teamwork Graph (TWG), autonomous Rovo agents, an Atlassian Model Context Protocol (MCP) server, and SHA-256 cryptographic proofs to mathematically guarantee release safety before code ever hits production.


What It Does

  • Causal Blast-Radius Simulator: Scans active Jira links, subtasks, and Confluence compliance policies in real time to calculate a deterministic Ship Confidence Score (0% to 100%) across multiple risk vectors.
  • Interactive Teamwork Graph Explorer: Dynamically traverses multi-hop organizational topologies (SCRUM-1 Core-Pay API ➔ SCRUM-2 Mobile Checkout ➔ POL-402 Confluence Policy ➔ Team Approvers) to expose hidden release breakers.
  • Autonomous Rovo Agent Silo Alignment: With one click, an assignable Rovo Agent (chronos-release-guardian) runs native Rovo actions, auto-provisions Jira remediation subtasks (SCRUM-5), alerts dependent teams, and elevates the Ship Confidence Score from 38% (Critical Risk) to 98% (Cleared for Release).
  • Bi-Directional Atlassian MCP Server: A native Forge WebTrigger exposing a spec-compliant JSON-RPC 2.0 MCP server, empowering external AI IDEs (Cursor, Antigravity, Claude Desktop) to query live Jira and Confluence context directly from the developer's terminal.
  • Zero-Trust SHA-256 Release Ledger: Hashes issues, dependencies, and approver signatures into tamper-evident Merkle receipts stored in Atlassian Forge Key-Value Storage (@forge/kvs) for audit-ready compliance.

How We Built It

  • Atlassian Forge (Node.js 22 runtime): Custom UI issue panel and serverless backend resolvers.
  • Rovo Agents & Actions: rovo:agent and modular rovo:action declarations for automated issue analysis and remediation.
  • Teamwork Graph (TWG) Engine: Multi-hop causal graph traversal mapping cross-project Jira issue links, Confluence policies, and active user approvers.
  • MCP Gateway: Spec-compliant JSON-RPC 2.0 WebTrigger endpoint with zero-trust token isolation.
  • Cryptographic Engine: SHA-256 Merkle proofs persisted in tenant-isolated Atlassian Forge Key-Value Storage (@forge/kvs).

Challenges We Ran Into

  • Cross-Service Graph Resolution: Building a low-latency graph engine inside Forge serverless functions required careful batching and caching to traverse Jira issue links and Confluence content without hitting execution timeout limits.
  • Spec-Compliant MCP over WebTriggers: Implementing a clean JSON-RPC 2.0 Model Context Protocol endpoint on an Atlassian Forge WebTrigger required precise request/response formatting compatible with external MCP clients (Cursor, Claude Desktop).

Accomplishments That We're Proud Of

  • Achieving a seamless transition from 38% Critical Risk to 98% Cleared for Release via one-click autonomous Rovo agent remediation.
  • Building a fully functioning, public-facing bi-directional Atlassian MCP server directly within the Atlassian Forge platform.
  • Zero lint warnings (forge lint = 0 issues) and strict least-privilege enterprise OAuth/Forge permission scopes.

What We Learned

  • The power of Atlassian's Teamwork Graph in surfacing non-obvious human and technical dependencies across disparate organizational tools.
  • How assignable Rovo agents transform AI from simple conversational chatbots into active, action-oriented workflow automation engines.

What's Next for ChronosMCP

  • Adding automated CI/CD pipeline blocking plugins (GitHub Actions / Bitbucket Pipelines) that reject pull requests if the ChronosMCP Ship Confidence Score falls below 80%.
  • Expanding MCP tool actions to allow IDE agents to directly propose Confluence policy revisions.

🏆 Codegeist 2026 Bonus Track Qualifications

  1. Best Agents (Rovo / Assignable Agents): Assignable Rovo Agent (chronos-release-guardian) equipped with 3 native actions that automatically analyzes issues and creates remediation subtasks.
  2. Best App with Atlassian MCP: Live JSON-RPC 2.0 WebTrigger MCP server allowing developers to query Jira context directly from AI IDEs.
  3. Best Apps with Teamwork Graph Integration: Multi-hop causal graph explorer visually connecting cross-project Jira issues, Confluence policies, and engineers.
  4. Best Marketplace Published App: Fully deployed to production with public installation capability, least-privilege scopes, and zero lint warnings.

Built With

Share this project:

Updates

Submission history