25 min

Build · A Slack channel digest bot

A bot that posts a daily summary to a Slack channel. Real API tokens, real scheduling. ~25 min.

You'll build a real Slack bot that posts a daily channel summary. Different shape than the existing builds: this one uses two API tokens (Slack + an LLM), reads real channel data, and gets scheduled.

0 of 9 steps

Create a Slack app + bot user

Once-per-bot setup. Skip if you've done this before.

Slack API: Your apps

api.slack.com/apps

create new

Create New App → From scratch. Name it (e.g. "Daily Digest").
Pick the workspace you want it in. Apps can be created in a personal workspace first; install in others later.
Under OAuth & Permissions, add these Bot Token Scopes:
- channels:history — read public channel messages
- chat:write — post messages
Click Install to Workspace. Approve.

You should see

You see a Bot User OAuth Token starting with xoxb-. Copy it.

Add the bot to a channel

The bot can only read channels it's a member of. In your test channel, type /invite @Daily Digest (or whatever you named it).

You should see

Slack confirms: "added Daily Digest to #channel".

Get the channel ID

In Slack, right-click the channel name → View channel details. The ID at the bottom looks like C0123ABC456.

Make a project folder + .env file

Shell

mkdir ~/slack-digest && cd ~/slack-digest
cat > .env <<'EOF'
SLACK_BOT_TOKEN=xoxb-...your-token...
ANTHROPIC_API_KEY=sk-ant-...
EOF
echo ".env" > .gitignore

You should see

Folder exists. .env is gitignored (never commit it).

Open your coding CLI and paste the build prompt

Run claude / codex / aider in the folder. Paste:

Build prompt

Write a Node.js script bot.js that:

- Reads SLACK_BOT_TOKEN and ANTHROPIC_API_KEY from env (.env via dotenv).
- Fetches the last 24h of messages from a channel ID passed as arg.
  - Use https://slack.com/api/conversations.history with limit=200, oldest=<24h ago>
  - Auth via Authorization: Bearer ${SLACK_BOT_TOKEN}
- Calls Claude with the messages joined by newlines, system prompt:
    "You are a TL;DR bot. Read this Slack channel transcript and produce a 3-bullet summary suitable
     for posting back to the channel. Each bullet starts with an emoji. Keep it under 80 words."
- Posts the summary to the same channel via chat.postMessage.

Single file. dotenv + node:18 fetch. No frameworks. Print every Slack API response so I can see failures.

Then add a package.json with a "start" script. Run with: node bot.js C123ABC456 (a channel id).

Approve the writes and the install

The CLI will:

Write bot.js (~80 lines)
Write package.json with the @anthropic-ai/sdk + dotenv deps
Run npm install — allow it

You should see

Done in ~30 seconds. node bot.js C0123ABC456 runs without crashing.

Test it

Run on your test channel:

Shell

node bot.js C0123ABC456

You should see

The bot posts a 3-bullet summary to the channel within a few seconds. If you see an error, paste it back to the CLI — it'll diagnose. Common gotchas: not_in_channel (re-invite the bot), missing_scope (check OAuth scopes).

Add --dry-run

Iterate in the CLI:

Iteration 1

Add a --dry-run flag. When set, print the summary to stdout instead of posting it. Useful for testing before pointing it at a real channel.

You should see

node bot.js C0123ABC456 --dry-run prints to stdout, doesn't post.

Schedule it (pick one path)

Ask the CLI for the three scheduling options:

Iteration 2

Show me three ways to schedule this for daily 9am runs:
1. macOS launchd (plist)
2. Linux cron (crontab line)
3. A GitHub Actions workflow that runs on schedule and pipes the SLACK_BOT_TOKEN from a secret

For each, show the exact file or command I'd run.

Recommended: GitHub Actions if you want zero machines to manage; cron if you have a server already; launchd if you only need it while your Mac is on.

You should see

The bot posts a daily summary at 9am without you doing anything.

What you just learned. Real bots are 80% glue code (API tokens, env vars, scheduling) and 20% AI. The hard part is the surface area — Slack's API, Anthropic's API, wherever you deploy. The LLM call itself is a one-liner. Apply this template to any SaaS-with-API (Linear, Notion, Discord, Telegram, Email) — same shape every time.