# Slack AI connector guide

Connect Slack data to AI agents with Combined. Explore users, channels, channel members, source setup, MCP access and read-only business queries.

Combined offers 1,125 connectors across catalog sources and connectors generated through desktop automation.

## Connect

1. Open Combined Connected Apps and select Slack.
2. Authorize the source and choose the relevant datasets.
3. Complete a first sync, then grant your agent access through the Set up or Access workspace.
4. Discover the actual source schemas before querying.

## Setup reference

- Channel name filter: optional
- Channel messages date window size (in days): optional
- Authentication mechanism: optional; Sign in via Slack (OAuth) or Bot Token
- Access token: depends on the selected authentication/setup option
- Client ID: depends on the selected authentication/setup option
- Client Secret: depends on the selected authentication/setup option
- Bot Token: depends on the selected authentication/setup option
- Include archived channels: optional
- Include private channels: optional
- Join all channels: required in the upstream source schema
- Threads Lookback window (Days): required in the upstream source schema
- Number of concurrent threads: optional
- Start Date: required in the upstream source schema
- Ignore messages with no replies in threads stream: optional

## Dataset reference

- users; key: id
- channels; key: id
- channel_members; key: member_id, channel_id
- channel_messages; key: channel_id, ts; update field: float_ts
- threads; key: channel_id, ts; update field: float_ts

This is the upstream connector reference, version 3.2.18. Account permissions and sync configuration determine the datasets available in your workspace.

## Agent prompt

Use Combined to analyze my Slack data.
List my granted sources, discover the datasets and describe their fields.
Look for datasets corresponding to users, channels, channel_members, channel_messages, threads.
Check the last successful sync and report the available date coverage.
Propose three useful questions based on the fields actually available.
For the question I choose, run read-only SQL over the complete matching dataset.
Use a bounded result; include source names, time window and query receipt.
Ask for a missing join key or metric definition instead of inventing one.

## Links

- Combined: https://www.trycombined.com/
- Setup: https://www.trycombined.com/docs/getting-started/onboarding
- MCP: https://www.trycombined.com/docs/integrations/mcp
- Pricing: https://www.trycombined.com/pricing
- Upstream source reference: https://docs.airbyte.com/integrations/sources/slack
- Full guide: https://www.trycombined.com/connectors/slack