Slack Agent Builder Challenge Submission

The Unified AI Context
Layer for Slack

One core engine. Three winning tracks. Nexus watches, remembers, and acts on your threads—turning scattered conversations into structured institutional memory.

Agent Builder

Native workflows

MCP Integrated

Secure tool access

RTS API

Live semantic search

What is Nexus?

Nexus is a single, powerful AI agent architecture designed to solve three completely different enterprise challenges. Instead of building a basic chatbot, we built a Unified Context Layer that connects Slack directly to external tools using the Model Context Protocol (MCP).

1. Nexus Flow

Built for Engineering. It tracks technical decisions, turns chat into Jira tickets, and searches GitHub docs seamlessly in Slack.

2. Nexus Impact

Built for Non-profits. It orchestrates disaster relief by assigning volunteers and tracking resources across chaotic emergency channels.

3. Nexus Enterprise

Built for IT Admins. It securely provisions new workspaces and monitors AI agent usage across the entire organization.

One Engine. Three Apps.

Nexus uses a single core architecture instantiated into three legitimately distinct submissions, targeting every track of the hackathon.

New Slack Agent

Nexus Flow

For Engineering & Product teams. Captures decisions from threads, auto-creates Jira/Linear tickets, and uses RTS to prevent duplicate questions before they're even asked.

MCP Tools Used

create_taskfetch_statussearch_kb
Slack Agent for Good

Nexus Impact

For Non-profits & Disaster Relief. Turns multi-channel chaos into assigned tasks and live resource ledgers. Uses cross-channel RTS to detect conflicting instructions.

MCP Tools Used

assign_volunteerlog_resource
Slack Agent for Orgs

Nexus Enterprise

For IT / Operations Admins. The same engine, packaged for org-wide install. Features org-ready RTS configuration, usage dashboards, and live Marketplace submission.

MCP Tools Used

provision_workspacemanage_billing

Core Architecture

How the required Slack technologies stack up in the Nexus engine.

Agent Reasoning Layer

The LLM core wrapped in a Bolt app. It decides per-turn whether to answer directly, call the MCP server for a tool, or search the workspace via the RTS API.

Nexus MCP Server

A custom Model Context Protocol server exposing track-specific tools (e.g. create_task, assign_volunteer, provision_workspace) with strict authorization checks.

Real-Time Search (RTS)

Executes semantic search strictly within the requesting user's permission boundary (`search:read.public`). No customer data leaks externally.

Data Flow (Single Turn)

1
User @mentions Nexus in a thread
2
Bolt routes event to Reasoning Layer
3
Reasoning Layer queries RTS API for context
4
Reasoning Layer invokes MCP tools if needed
5
Nexus replies with Block Kit UI

Enterprise Admin Console

Govern agent usage, monitor MCP tool calls, and control org-ready RTS permissions across all connected workspaces.

1
Active Workspaces
0
Real-time Queries
0
MCP Invocations
0
Data Leakage

Live MCP Tool Audit Log

Waiting for live tool invocations from Slack...

RTS Engine Status

Index FreshnessReal-time
Search Scope EnforcedActive
Queries strictly bound to caller's search:read.public capability via OAuth.