Inspiration
In high-security enterprise environments, Slack AI, Notion AI, bots, plugins, and API access are all disabled. Teams were manually digging through noisy Slack channels every day, copying and pasting customer issues into Notion. This bottleneck inspired us to build something that works even when everything is locked down: an xAI-powered orchestrator that automates the workflow without needing any privileged access.
What it does
The Groklet reads customer conversations inside Slack, summarizes them using xAI, and automatically creates structured, reviewable issues inside Notion. It works with zero Slack bot tokens, zero Notion API usage, and no browser extensions. It mimics human actions through RPA, allowing fully automated cross-app workflows in environments where traditional automation is forbidden.
How we built it
We implemented an RPA-first architecture using Playwright to emulate human browser behavior. A central backend orchestrator powered by xAI coordinates every step: collecting Slack messages, extracting context, summarizing threads, generating issue drafts, and posting them into Notion. All reasoning is handled with privacy alignment: no logs, no retention, and encrypted inference.
Challenges we ran into
- No Slack bot tokens allowed
- No Notion API access permitted
- No MCP plugins or automation extensions
- UI inconsistencies across Slack and Notion
- Building a robust RPA layer that behaves like a human
- Ensuring reliability despite unpredictable DOM changes
Accomplishments that we're proud of
- Automating a workflow that was previously considered “impossible” without APIs
- Building a universal cross-app automation layer using only browser actions
- Producing consistent, high-quality issue summaries using xAI
- Creating a reusable foundation that can integrate Slack, Notion, Linear, Confluence, Zendesk, and more
- Demonstrating a privacy-aligned AI orchestrator that respects enterprise constraints
What we learned
- RPA becomes extremely powerful when paired with an LLM that can maintain workflow state
- Enterprise restrictions force creative designs that break away from API-centric thinking
- Cross-app orchestration needs a single “brain” to unify context
- Even noisy, unstructured Slack conversations can be transformed into clean, actionable units with the right pipeline
What's next for The Groklet
- Support for additional tools like Linear, Confluence, Jira, and Zendesk
- Adding memory, Q&A, and retrieval layers to turn it into a fully agentic workflow assistant
- Improving the RPA engine with multi-step reasoning and UI adaptation
- End-to-end workspace intelligence: from Slack threads to dashboards, metrics, insights, and alerts
- Eventually turning The Groklet into a universal enterprise automation layer that requires zero integration permissions
Built With
- docker
- fastapi
- humanemulator
- notion
- playwright
- postgresql
- python
- rpa
- slack
- xai
Log in or sign up for Devpost to join the conversation.