DevSync AI

Inspiration

Every software team knows the pain of keeping everyone on the same page. Developers spend time updating tickets, project managers ask for progress summaries, and important information often gets buried in GitHub comments or Slack threads. We were inspired by this constant overhead. Instead of humans doing all the context-switching, why not build an AI-powered release coordination tool that automates it? That was the starting point for DevSync AI: a system that listens to development activity across GitHub, JIRA, and Slack, and then automatically generates the updates that teams usually have to manually prepare.

What We Learned

Working on DevSync AI taught us a lot about how to combine AI-assisted coding with real-world developer workflows. • We learned how to use Kiro’s spec-to-code approach: writing high-level natural language specs like “fetch all open PRs and send a daily Slack summary,” and letting Kiro generate functional backend code. • We deepened our understanding of multi-service integration — connecting GitHub APIs, JIRA APIs, and Slack webhooks to create a unified workflow. • We explored how to structure agent hooks that run on triggers (PR creation, ticket updates, weekly changelogs). • We also discovered how important it is to guide AI carefully — the more precise the spec and prompt, the better the generated code.

How We Built It

We approached the build in phases, guided by a clear two-week roadmap: 1. Foundation: We set up the repo, Supabase backend, and initial FastAPI infrastructure using Kiro. 2. GitHub Integration: We connected to the GitHub API, fetching pull requests, summarizing open/merged ones, and checking for conflicts. 3. Slack Agent Hook: We configured a webhook so that PR activity could automatically post summaries into Slack channels in a clean, readable format. 4. JIRA Integration: We extended the system to track ticket updates and sprint progress, syncing that into both Slack and an optional dashboard. 5. Automation Hooks: We created scheduled jobs — for example, a Friday changelog generator that pulls from commits, PRs, and issues, then delivers a Markdown summary. 6. Advanced Features: We experimented with blocker detection — monitoring for inactivity, unresolved conflicts, or duplicated work — and notifying project managers proactively. 7. Polish & Deployment: We focused on error handling, reliability, and preparing a demo-ready version of the tool, including a short video walk-through.

Throughout, we used Kiro as a coding partner — sometimes scaffolding entire modules, other times refining snippets or generating quick integrations via conversational prompts.

Challenges We Faced

Building DevSync AI was exciting but also full of hurdles: • API Complexity: Each service (GitHub, JIRA, Slack) had different authentication methods, rate limits, and data formats. Orchestrating them in a single backend was tricky. • Notification Balance: We didn’t want Slack updates to become spam. Finding the right structure and frequency of messages was a design challenge. • Debugging AI-Generated Code: While Kiro sped things up, the generated code sometimes needed adjustments. Learning how to “prompt engineer” for reliable output was a skill in itself. • Team Coordination: Since we worked in parallel on different integrations, syncing progress and merging our codebases smoothly required discipline. • Deployment & Testing: Setting up the environment, secrets, and ensuring reliability across integrations was harder than expected.

Built With

• kiro