SignalGuard

SignalGuard SignalGuard is a Slack agent that helps people protect their deep work time by reducing notification fatigue and surfacing only what actually matters. Most teams face two major problems in Slack: constant notifications that break focus, and the difficulty of finding important context buried in noisy channels. SignalGuard solves both by combining Slack’s Real-Time Search (RTS) API with Model Context Protocol (MCP) and intelligent reasoning. When a user says “Protect my focus for 2 hours”, the agent uses RTS to intelligently scan recent conversations for urgent mentions, blockers, and action items. It then uses an LLM-powered reasoning layer to categorize information and takes meaningful actions using MCP — including creating structured summaries, updating user status, and organizing information into clean threads. The result is a calmer, more productive Slack experience where people can focus deeply while still staying on top of what’s important. How we built it We built SignalGuard using Slack’s official Bolt framework and the slack create agent template. The core intelligence comes from a reasoning loop that combines:

Real-Time Search (RTS) API to fetch relevant, permission-respecting context from channels and threads Model Context Protocol (MCP) to securely execute actions such as creating summaries and managing user status LLM reasoning (using Groq) to intelligently categorize information instead of relying only on keyword rules

All interactions happen naturally inside Slack using threads, Block Kit components, and status updates for a seamless experience. Challenges & Learnings One challenge was balancing urgency detection — early versions either missed important items or surfaced too much noise. We iterated on the reasoning prompt and RTS strategy until the agent felt helpful rather than overwhelming. We also learned how powerful RTS and MCP become when combined with LLM reasoning. This combination allows agents to not just answer questions, but actively protect user focus and reduce cognitive load.

Built With

• groq-cloud
• mcp
• rts-api