Slack Login

Slack Login

Instructions

1. Check prerequisites

1. Verify Python 3.11+ is available: python3 --version
2. Check if the playwright pip package is installed: python3 -c "import playwright". If it fails, install it: pip3 install playwright
3. Check if Chromium is installed for Playwright: python3 -c "from playwright.sync_api import sync_playwright; p = sync_playwright().start(); b = p.chromium.launch(headless=True); b.close(); p.stop()". If it fails, run: playwright install chromium

2. Determine workspace URL

Parse the user's prompt for a workspace URL or name:

• If the prompt contains a full URL (e.g., https://<YOUR_SLACK_WORKSPACE_URL>), use that URL.
• If the prompt contains a workspace name but not a full URL, construct it: https://<name>.slack.com
• Otherwise default to https://app.slack.com/client/

3. Run the extraction script

Run the bundled script with a long Bash timeout (600 seconds / 10 minutes) to allow for slow SSO/MFA logins:

python3 scripts/slack-login.py --workspace <URL> --json --output ~/.slack_tokens.env --timeout 300

The script path is relative to this skill's directory.

Before running, tell the user:

A browser window is about to open. Log in to your Slack workspace there, wait until the workspace fully loads (you see channels and messages), then come back here. The script will automatically detect when you're logged in and extract the tokens.

Important: The script prints status messages to stderr and the JSON result to stdout. Parse the JSON from stdout.

4. Parse the result

• If the script exits 0, parse the JSON from stdout. It contains xoxc_token, xoxd_token, and team_id.
• If the script fails (non-zero exit), show the stderr output to the user and suggest:
  • Make sure they fully logged in and waited for the workspace to load
  • Try running again — the browser profile is persistent so SSO may be faster next time
  • Check network/VPN connectivity

5. Configure the Slack MCP server

5a. Configure Claude Code MCP (~/.claude/mcp.json)

Read ~/.claude/mcp.json (create it if it doesn't exist — start with {"mcpServers": {}}).

Add or update the slack entry under mcpServers:

{
  "mcpServers": {
    "slack": {
      "command": "<HOME>/slack-mcp/.venv/bin/python",
      "args": ["slack_mcp_server.py"],
      "cwd": "<HOME>/slack-mcp",
      "env": {
        "SLACK_XOXC_TOKEN": "<xoxc_token>",
        "SLACK_XOXD_TOKEN": "<xoxd_token>"
      }
    }
  }
}

Replace <HOME> with the actual home directory path, and fill in the extracted tokens.

5b. Configure Cursor MCP (.cursor/mcp.json in project root)

Read .cursor/mcp.json in the project root (create it if it doesn't exist — start with {"mcpServers": {}}).

Add or update the slack entry under mcpServers with the same structure as above.

6. Verify and report

Tell the user:

1. What was configured (show the paths of both MCP config files updated)
2. That they should restart their editor or reload MCP servers for the changes to take effect
3. Suggest testing with: "Try asking me to use the slack whoami tool to verify the connection works"

Example prompts

• "Slack login"
• "extract Slack tokens"
• "set up Slack MCP"
• "Slack login https://"
• "refresh my Slack tokens"