Neuron

An AI-powered Slack agent that builds your engineering team's memory graph from GitHub and Docs while using live Slack context to answer and automate work.

Comment

Neuron — The Engineering Memory Graph for Slack

Inspiration

Engineering knowledge is everywhere—but rarely in one place.

A design decision starts in a Slack thread, evolves into a GitHub discussion, becomes a pull request, and finally gets documented in Google Docs or an RFC. Months later, when someone asks "Why did we build it this way?", the answer is buried across multiple tools.

Current AI assistants can search documents, but they don't truly understand the relationships between conversations, code, documentation, and decisions.

We wanted to build something that acts like an experienced engineer on the team—not just answering questions, but remembering why decisions were made and connecting the entire engineering workflow.

That idea became Neuron: an AI-powered engineering memory built natively for Slack.

What it does

Neuron is a Slack-native AI agent that builds a living engineering memory for your organization.

It connects with engineering tools such as GitHub and Google Docs to build a semantic knowledge graph using Cognee Cloud, while using Slack's Real-Time Search (RTS) API to retrieve live workspace context whenever it's needed.

With Neuron, engineers can:

  • 🧠 Ask architecture and engineering questions directly in Slack
  • 🔍 Discover why technical decisions were made
  • 📄 Generate documentation and onboarding guides
  • 🐛 Create GitHub issues without leaving Slack
  • 💬 Summarize Slack discussions and GitHub threads
  • 🔗 Connect discussions, pull requests, issues, and documentation together
  • ⚡ Find the right engineer, service, or historical context in seconds

Instead of searching across multiple tools, engineers interact with a single AI assistant that understands the relationships between their knowledge.

How we built it

Neuron is being built entirely in TypeScript, allowing us to use one language across the entire stack.

The project combines several modern technologies:

Slack

  • Slack Agent SDK
  • Slack Agent Builder
  • Slash Commands
  • App Mentions
  • Block Kit
  • Real-Time Search (RTS) API

Knowledge Layer

  • Cognee Cloud
  • Knowledge Graph
  • Semantic Search
  • Relationship Extraction

Integrations

  • GitHub API
  • Google Docs API
  • Gemini for reasoning and planning
  • Model Context Protocol (MCP) architecture for future integrations

The overall architecture separates knowledge into two layers:

  • Persistent Memory – GitHub repositories, pull requests, issues, documentation, and engineering artifacts are stored in a semantic knowledge graph.
  • Live Context – Slack conversations remain inside Slack and are retrieved on demand using RTS. Neuron extracts engineering knowledge from relevant discussions without storing raw Slack messages.

This hybrid architecture allows the agent to combine long-term organizational memory with fresh workspace context.

Challenges we ran into

1. Designing the Memory Graph

The biggest challenge was designing the knowledge graph itself.

Engineering knowledge isn't just a collection of documents—it is a network of relationships.

We spent considerable time deciding:

  • Which entities should exist
  • How they should be connected
  • What belongs in long-term memory
  • What should remain transient

Finding the right balance between persistent knowledge and live context was one of the most difficult architectural decisions.

2. Connecting Multiple Knowledge Sources

GitHub, Google Docs, and Slack all expose information differently.

Repositories contain structured engineering data, documentation contains long-form knowledge, while Slack discussions are dynamic and constantly changing.

Creating a unified reasoning system capable of understanding all three without duplicating data required careful planning around ingestion, entity extraction, and relationship mapping.

Accomplishments that we're proud of

Although Neuron is still under active development, we're proud of the foundation we've built.

So far we've successfully:

  • ✅ Designed a scalable architecture around a semantic knowledge graph
  • ✅ Built the initial GitHub integration for engineering knowledge ingestion
  • ✅ Created a modular TypeScript codebase designed for future MCP integrations
  • ✅ Defined a hybrid memory architecture combining persistent knowledge with live Slack context
  • ✅ Planned the project with production-ready documentation, security, and roadmap from day one

Building a strong architectural foundation now will make it much easier to expand Neuron into a full engineering assistant.

What we learned

This project taught us that organizational memory is fundamentally a graph problem, not a search problem.

The most valuable engineering knowledge isn't contained within individual documents—it's hidden in the relationships between discussions, code changes, documentation, and decisions.

We also learned how powerful Slack's AI platform can be when combined with external knowledge systems.

Using Slack's Real-Time Search API together with a semantic knowledge graph allows Neuron to answer questions using both historical engineering knowledge and the latest workspace discussions.

Finally, we gained a much deeper understanding of agentic AI architectures, knowledge graphs, and designing systems that reason instead of simply retrieving information.

What's next for Neuron

Our vision extends far beyond this hackathon.

The next milestones include:

  • 🚀 Complete Google Docs integration
  • 🚀 Automatic knowledge graph updates from GitHub webhooks
  • 🚀 Native MCP server integrations
  • 🚀 Visual knowledge graph explorer
  • 🚀 Architecture decision timeline
  • 🚀 Engineering ownership detection
  • 🚀 Automatic onboarding document generation
  • 🚀 Cross-repository reasoning
  • 🚀 Slack Marketplace submission
  • 🚀 Support for Jira, Notion, Confluence, and Linear

Ultimately, we want Neuron to become the organizational memory layer for engineering teams—a system that continuously learns, preserves institutional knowledge, and helps every engineer understand not just what was built, but why it was built.

Knowledge shouldn't disappear into chat history. Neuron transforms engineering knowledge into a living organizational memory that grows with your team.

Built With

Share this project:

Updates