Channel Concierge

• 

  Architecture: Bolt app over Socket Mode calling Slack's Real-Time Search API directly — zero external LLM or paid search service.

• 

  Channel Concierge in action — a real question gets routed to three existing threads with matching context.

Inspiration

In any growing Slack workspace, the same questions get asked over and over — because the answer already exists, buried in some thread nobody can find. New hires re-ask what veterans already solved. Teams duplicate work because they didn't know another channel had already worked through it. We wanted to fix that at the exact moment the question is typed.

What it does

Channel Concierge is a Slack agent built entirely on Slack's native Real-Time Search API. Mention the bot with a question, and it searches your workspace in real time — messages and channels you have access to — and routes you to the existing conversation with your answer, complete with direct "Jump to thread" links.

How we built it

Node.js and the Slack Bolt SDK, running over Socket Mode — no public server or webhook infrastructure required. When a user @-mentions the bot, Slack issues a short-lived action token. We use that token to call assistant.search.context — the Real-Time Search API — directly, scoped to exactly what the requesting user can see. Results are formatted into a threaded Block Kit reply with clickable links to the source conversation.

Challenges we ran into

Getting the Real-Time Search API wired up correctly took real debugging — from figuring out the correct OAuth scope (search:read.public), to discovering the action token is only issued on mention/DM events rather than ambient channel messages, to matching the actual nested response shape of the API. None of this was obvious from a first read of the docs.

Accomplishments that we're proud of

A fully working, zero-external-cost Slack agent — no LLM API, no third-party search service, no paid hosting. Just Slack's own platform primitives, solving a problem every growing team has.

What we learned

How Slack's Real-Time Search API actually works under the hood — the action token lifecycle, scope requirements, and response format — none of which is obvious from the docs alone. Also a deeper appreciation for how much can be built entirely on native platform primitives without reaching for external services.

What's next for Channel Concierge

Expanding to DMs and the Assistant panel, and adding smarter ranking so results factor in recency and channel activity, not just keyword relevance.

Built With

• javascript
• node.js
• slack-bolt-sdk
• slack-real-time-search-api
• socket-mode