slackcli

For AI, by AI. The Slack CLI and MCP server built from the ground up for LLM-powered agents and programmatic access.

Why slackcli?

Most Slack tools are built for humans. slackcli is built for agents. Every design decision — JSON-first output, structured error codes, automatic pagination, rate limit retries — is made so LLMs can interact with Slack reliably without hand-holding. Plug it in as an MCP server and your AI agent gets Slack superpowers. Use --read-only mode when you want guardrails.

Quick Start

1. Create a Slack app and install it to your workspace with the scopes you need (e.g. channels:read, chat:write, search:read).
2. Grab the Bot or User OAuth Token (xoxb-...) from OAuth & Permissions.
3. Run:

brew install jackchuka/tap/slackcli
export SLACK_TOKEN=xoxb-your-token
slackcli channels list

Features

• CLI commands for channels, messages, users, files, reactions, and search
• MCP server (stdio transport) for AI agent integration
• JSON-first output optimized for LLM consumption, with TTY-aware table fallback
• Rate limit handling with automatic retry
• Error classification with structured error codes
• Pagination support across all list operations
• Read-only mode to prevent accidental writes by AI agents

Installation

Homebrew:

brew install jackchuka/tap/slackcli

Or via Go:

go install github.com/jackchuka/slackcli/cmd/slackcli@latest

Or build from source:

make build

Authentication

Set your Slack token via environment variable or the CLI:

# Environment variable
export SLACK_TOKEN=xoxb-your-token

# Or use the auth command
slackcli auth login --token xoxb-your-token

# With a named workspace
slackcli auth login --token xoxb-your-token --name my-workspace

Token resolution order: --token flag > SLACK_TOKEN env > stored config.

Usage

CLI

# Channels
slackcli channels list
slackcli channels info C1234567890
slackcli channels create new-channel

# Messages
slackcli messages list --channel C1234567890
slackcli messages send --channel C1234567890 --text "Hello"
slackcli messages search --query "important"

# Users
slackcli users list
slackcli users info U1234567890

# Files
slackcli files list
slackcli files upload --channel C1234567890 --file ./report.pdf

# Reactions
slackcli reactions add --channel C1234567890 --timestamp 1234567890.123456 --name thumbsup
slackcli reactions list --user U1234567890

MCP Server

Start the MCP server for AI agent integration:

slackcli mcp serve

Configure in your MCP client (e.g., Claude Desktop):

{
  "mcpServers": {
    "slack": {
      "command": "slackcli",
      "args": ["mcp", "serve"],
      "env": {
        "SLACK_TOKEN": "xoxb-your-token"
      }
    }
  }
}

Available MCP tools:

Read-Only Mode

Use --read-only to restrict to read-only operations. This prevents AI agents from accidentally sending messages, deleting files, or modifying channels.

CLI -- write commands are rejected before execution:

slackcli --read-only channels list           # works
slackcli --read-only messages send ...       # blocked
# Error: command "slackcli messages send" is a write operation and cannot be used in read-only mode

MCP server -- write tools are hidden entirely (AI agents never see them):

{
  "mcpServers": {
    "slack-readonly": {
      "command": "slackcli",
      "args": ["--read-only", "mcp", "serve"],
      "env": {
        "SLACK_TOKEN": "xoxb-your-token"
      }
    }
  }
}

Read-only tools (always available): auth_test, list_channels, get_channel_info, list_messages, list_users, get_user_info, get_user_presence, list_reactions, list_files, get_file_info, search_messages.

Write tools (hidden in read-only mode): create_channel, archive_channel, invite_to_channel, kick_from_channel, set_channel_topic, set_channel_purpose, send_message, edit_message, delete_message, add_reaction, remove_reaction, delete_file.

Output Formats

# Force JSON output
slackcli channels list -o json

# Force table output
slackcli channels list -o table

Default: table for TTY, JSON for piped output.

License

MIT