SlackScope — AI Project Health Dashboard for Slack

SlackScope is an intelligent Slack agent that monitors your project channels and provides real-time health dashboards. It uses AI to detect blockers, analyze team sentiment, and surface risks

Comment

Inspiration

Blockers get buried in Slack. Managers find out too late. We built an always-on agent that flags risks before they become crises.

What it does

  • /slackscope health — AI health score (0-100), sentiment, blockers with severity, recommendations. Rich Block Kit cards.
  • /slackscope team — Contribution trends, collaboration score, peak hours, unresolved questions.
  • /slackscope report — Multi-channel dashboard, worst-first. One glance = full portfolio status.
  • /slackscope trends — Historical scores with directional trend arrows.
  • @SlackScope — Natural language Q&A about any project channel.
  • Smart Alerts — 5 rules (critical drops, blocker pileups, score nosedives, negative sentiment, inactivity). Auto-DMs with cooldown.
  • MCP Server — 5 tools for external AI agent integration.
  • Interactive UI — Deep Dive, View Blockers, Snooze buttons. Config modal for channel selection.

How we built it

Python + Slack Bolt (Socket Mode), Groq Llama 3.3 70B (free, sub-second), Slack RTS API with conversations.history fallback, FastMCP, APScheduler, JSON persistence. Zero database. Hosted on Railway.

Challenges we ran into

  • RTS API sandbox limits → robust fallback + graceful degradation
  • Parsing casual Slack chaos (typos, emojis, threads) into structured JSON → precise prompts + response_format enforcement
  • Alert fatigue → per-channel per-type cooldowns with persistent state
  • Slack's 3-second ack() deadline with 1-2s LLM calls

Accomplishments that we're proud of

  • $0 cost — Groq free + Railway free + Socket Mode (no public URL)
  • Fully native Slack feel — Block Kit cards, modals, interactive buttons
  • 5 smart alert rules with intelligent cooldown
  • MCP integration — composable for larger AI workflows
  • Clean architecture — agents/, services/, slack_handlers/, ui/, mcp_server/

What we learned

  • Block Kit can feel first-party with score bars, multi-field layouts, action rows
  • Structured LLM output needs schema prompts + response_format + defensive fallbacks
  • Alert engine was harder than analyzer — cooldowns, severity rules, state persistence
  • Socket Mode = hackathon game-changer. No URL management, instant dev

What's next for SlackScope — AI Project Health Dashboard for Slack

Jira + GitHub integration, auto-generated standups, trend prediction, per-user dashboards.

Built With

Share this project:

Updates

Submission history