Skip to main content

Duolingo AI Slackbot – Aaron Wang

Speaker: Aaron Wang, Software Engineer (DevXAI), Duolingo

Journey Timeline

DateMilestone
Nov 2024MCP introduced, engineers start experimenting
May 2025Centralized "MCP store" with setup instructions
Aug 2025MCP standardization efforts begin
Sep 2025AI Slackbot launches
Apr 2026250+ weekly users (30% of company), 80% approval rate

Architecture

  • Claude Code Agent SDK — AI brain
  • Slack Bot API — messaging interface
  • 50+ MCP servers connected
  • Tools: AWS, BigQuery, internal services, external APIs

Standardization Strategy

Server TypeApproach
First-party (Linear, etc.)Connect directly from AI client
Open-sourceFork, add auth, host internally via HTTP
Shared credentials (Jenkins)Shared service token on internal server
Individual credentials (Google, Slack)Run on internal service VM
Internal servicesPython library to convert to MCP, host internally
IAM-controlledUse add tools, not MCP
Local-only (Playwright)Support via stdio

Security Principles

  1. Bot is not a way to bypass permissions — different access layers by group
  2. No write operations without human approval — only Slack responses autonomous
  3. Sandboxed environment — no access to other machine resources
  4. No side channel between users — messages sandboxed per-person

Key Features

  • Respond in channels + DMs
  • Group verification for write operations
  • System prompts for internal info sources
  • Channel-specific documentation
  • Power skills and starter agents
  • Feedback collection + regression tests

Open Source

Repository: github.com/duolingo/slack-ai-agents

Key Quote

"Even a single click to copy a config is still too much friction. We wanted to provision MCP servers to people automatically."