Slack - Copilotkit — ContextQMD

Connect Slack yourself with the Slack adapter. It runs over Socket Mode by default, so no public URL is needed for local development.

Install

bash

npm install @copilotkit/channels @copilotkit/runtime @tanstack/ai @tanstack/ai-openai

Credentials

From your Slack app:

bash

SLACK_BOT_TOKEN=xoxb-...   # Bot User OAuth token
SLACK_APP_TOKEN=xapp-...   # App-level token, for Socket Mode

# The runtime runs every channel, so a CopilotKit Intelligence key is required
# even for a direct adapter (free tier available).
COPILOTKIT_API_KEY=cpk-...

Connect

import { createChannel } from "@copilotkit/channels";
import { slack, defaultSlackContext } from "@copilotkit/channels/slack";
import { CopilotRuntime, CopilotKitIntelligence } from "@copilotkit/runtime/v2";
import { createCopilotNodeListener } from "@copilotkit/runtime/v2/node";
import { agent } from "./agent"; // your TanStack AI agent, see the Quickstart

const channel = createChannel({
  name: "support-bot",
  agent,
  // Slack output-formatting guidance so replies render well in Slack.
  context: [...defaultSlackContext],
  adapters: [
    slack({
      botToken: process.env.SLACK_BOT_TOKEN!,
      appToken: process.env.SLACK_APP_TOKEN!,
      respondTo: {
        directMessages: true,
        appMentions: { reply: "thread" },
        threadReplies: "mentionsOnly",
      },
    }),
  ],
});

channel.onMessage(async ({ thread, message }) => {
  await thread.runAgent({ prompt: message.text });
});

// The runtime drives the channel — it starts the Slack adapter, you don't call
// channel.start() yourself. A CopilotKit Intelligence key runs every channel,
// direct adapters included (free tier available); `ready()` activates it.
const runtime = new CopilotRuntime({
  agents: {},
  intelligence: new CopilotKitIntelligence({
    // apiUrl and wsUrl default to the managed Intelligence platform. Override
    // both together only for a self-hosted deployment.
    apiKey: process.env.COPILOTKIT_API_KEY!,
  }),
  identifyUser: () => ({ id: "demo-user", name: "Demo User" }),
  channels: [channel],
});

const listener = createCopilotNodeListener({ runtime });
await listener.channels.ready();

respondTo controls when the bot answers: DMs, mentions, and whether it keeps listening in a thread it is part of.

Streaming

The Slack adapter streams replies as they are produced. The streaming option picks the transport:

"native" (default) uses Slack's chat.startStream in threads, and falls back automatically in flat DMs or workspaces where the streaming API is not available.
"legacy" edits a single message with repeated chat.update calls.

slack({
  botToken: process.env.SLACK_BOT_TOKEN!,
  appToken: process.env.SLACK_APP_TOKEN!,
  streaming: "native",
});

<Callout type="info" title="Declare slash commands in the manifest"> A `/command` only reaches your bot if it is declared in your Slack app manifest with the same name you pass to `defineChannelCommand`. See [commands and reactions](/channels/commands-and-reactions). </Callout>

Built-in tools and context

The Slack adapter ships helpers you can spread into your channel: a set of tools (including a Slack user-lookup tool) and context that guides Slack formatting and tagging.

import { slack, defaultSlackTools, defaultSlackContext } from "@copilotkit/channels/slack";

createChannel({
  name: "support-bot",
  agent,
  tools: [...defaultSlackTools],
  context: [...defaultSlackContext],
  adapters: [slack({ /* ...credentials */ })],
});