Your assistant inside the workspace — read channels, search history, post messages and tame the noise.
Slack is where decisions get made and then immediately buried. The Slack MCP server gives your assistant a line into the workspace — listing channels, reading history, posting messages, replying in threads and reacting — which turns "scroll back through three weeks of #engineering" into a question you can just ask.
Summarisation is the headline act. "What happened in #incidents this week?" produces a digest with the noise stripped out; "what did we decide about the pricing change?" pulls the relevant exchange from a channel you'd have skimmed past. For anyone returning from leave to four-digit unread counts, this alone justifies the setup. Posting works too — the assistant can draft an update in context and, with your approval, deliver it to the right channel or thread.
Setup goes through a Slack app: create one at api.slack.com, grant bot-token scopes (channels:read, channels:history, chat:write and friends), install it to the workspace and copy the xoxb- token into the server config. Ten minutes, mildly bureaucratic, entirely one-time. Note that the bot sees only channels it's been invited to — a natural scoping mechanism worth using deliberately rather than inviting it everywhere.
The original Anthropic reference implementation is archived but functional; community successors (korotovsky's slack-mcp-server notably) extend it with search modes, DM support and Enterprise workarounds. For most teams the reference server's read-summarise-post loop covers the daily value: fewer tabs, fewer scrollbacks, decisions surfaced instead of excavated.
channels:history only covers public channels. Private channels need groups:history plus an explicit invite, and DMs need im:history — miss a scope and the tool returns an empty result rather than a permissions error, which looks identical to a quiet channel.SLACK_CHANNEL_IDS as an allowlist even when the bot lives in only a handful of places. It keeps the channel listing small, which measurably reduces the assistant's tendency to wander.Condense a week of a busy channel into the five things that mattered.
Find where something was agreed and quote the actual exchange.
Compose status updates in context and post them after your approval.
| Tool | What it does |
|---|---|
| slack_list_channels | List channels the bot can see |
| slack_get_channel_history | Read recent messages from a channel |
| slack_post_message | Post a message to a channel |
| slack_reply_to_thread | Reply within a thread |
| slack_add_reaction | React to a message with an emoji |
| slack_get_thread_replies | Read a full thread |
| slack_get_users / slack_get_user_profile | Look up workspace members |
{
"mcpServers": {
"slack": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-slack"],
"env": {
"SLACK_BOT_TOKEN": "xoxb-your-bot-token",
"SLACK_TEAM_ID": "T01234567"
}
}
}
}Create a Slack app with bot scopes (channels:read, channels:history, chat:write, reactions:write, users:read), install to workspace, copy the bot token.
A Slack app with bot-token scopes installed to your workspace. The bot only sees channels it's invited to.
| Variable | Description | Required |
|---|---|---|
| SLACK_BOT_TOKEN | Bot user OAuth token from your Slack appe.g. xoxb-... | Yes |
| SLACK_TEAM_ID | Workspace/team IDe.g. T01234567 | Yes |
| SLACK_CHANNEL_IDS | Optional comma-separated channel allowlist | Optional |
Inbox intelligence — search, read, draft and send Gmail through your assistant with OAuth auto-setup.
Read and write Jira, Confluence, Bitbucket, JSM and Compass from your AI client — with your own permissions.
Read and send Telegram messages through your assistant — chats, channels and history via the client API.
Enables Discord bot integration with Model Context Protocol (MCP) compatible applications like Claude Desktop.
Exposes REST APIs defined by OpenAPI specifications as Model Context Protocol (MCP) tools, facilitating seamless integration into MCP-based workflows.
Enables AI assistants to interact with the Discord platform via a Model Context Protocol (MCP) server.