Sherpa Data Sherpa Data Book a demo
Sherp Beta

Ask your Slack workspace anything — and get an actual answer.

Sherp is a Slack bot that answers natural-language questions about your workspace. It searches your history, reads threads and any attached images, and synthesizes a plain-English answer with citations back to the source messages. Not a list of search results — a sherp.

Built for ops teams who know the answer is "somewhere in Slack" and don't want to spend 20 minutes scrolling to find it.

Sherp — the AI Slack Q&A mascot

What's inside

Smarter Slack search, with cited answers.

Natural-language, not keyword search

Ask like a human. Get an answer like a human.

You ask "what did we decide about the Oak Street key system?" — Sherp extracts the search queries a good researcher would run, merges the results across them, and writes you back an actual answer. Not a list of messages. A sherp.

  • Claude extracts 2–4 targeted search queries from your question
  • Results merged across queries, deduped, and re-ranked for relevance
  • Synthesized plain-language answer, not a list of search hits

Screenshot coming

Threads and images, not just one-liners

Reads the whole conversation. And the screenshots.

The answer's rarely in a single message. Sherp fetches thread replies for context and downloads + analyzes any images referenced — a screenshot of an error, a photo of a broken lock, a whiteboard. So the answer includes what's in them.

  • Thread expansion pulls full reply chains into context
  • Image download + vision analysis for screenshots, photos, whiteboards
  • Time-aware — "last two days" or "this week" auto-adds date filters

Screenshot coming

Citations, always

Every claim points back to the source message.

Sherp's answers come with [1][2][3] citations that link to the actual Slack messages. No hallucinated decisions, no made-up quotes — if it's in the answer, it came from a real message you can click through to.

  • Inline [N] citations with message author, channel, and date
  • Click-through to the exact Slack thread
  • If the answer isn't in your Slack, Sherp says so — no guessing

Screenshot coming

Per-user permissions

Your Sherp only searches what you can see.

Sherp uses per-user OAuth — each teammate authenticates individually and their Sherp searches are scoped to the channels they have access to. Private channel questions stay private; the ops team doesn't accidentally search exec DMs.

  • Per-user OAuth via Slack — searches respect each user's channel access
  • Ephemeral responses by default (only you see the answer)
  • Query log stored privately, not trained into any shared model

Screenshot coming

Multilingual

Ask in your language. Sherp reads and replies in it.

Sherp understands Slack messages in any language Claude handles and replies in whatever language you asked the question — so an operator in Spain can query a Slack workspace with English and Portuguese messages and get a clean Spanish answer.

  • Strongest fluency in English, Spanish, French, German, Italian, Portuguese, Dutch, Russian, Japanese, Korean, Chinese, Arabic
  • Reads messages in one language, answers in another without losing fidelity
  • 100+ languages supported via Claude Sonnet 4

Screenshot coming

Integrations

Slack-native, Claude-powered.

Sherp lives in Slack and only searches Slack. Your messages aren't shipped anywhere, and queries aren't used to train anyone's model.

  • Slack
  • Claude (Anthropic)
  • Socket Mode
  • Per-user OAuth

Beta — limited spots

Join the Sherp waitlist.

We're onboarding teams into the Sherp beta in small batches to tune search and synthesis against real Slack workspaces. Drop your email and tell us what you'd use Sherp for, and we'll reach out when there's a spot open.