Introducing ChatBotKit CLI Agent Mode

We're excited to announce a major enhancement to the ChatBotKit CLI - a powerful new agent mode that brings autonomous AI capabilities directly to your command line. Now you can use your ChatBotKit bots combined with local file system and command execution tools to automate development tasks, business processes, and much more.

What is Agent Mode?

Agent mode transforms the ChatBotKit CLI into an autonomous AI assistant that can interact with your local environment while leveraging the full power of your ChatBotKit platform resources. It bridges the gap between your local development environment and remote integrations, enabling powerful automation workflows that were previously complex to implement.

Key Features

Local File System Access

The CLI agent comes with built-in tools for interacting with your local file system:

read - Read file contents
write - Write content to files
edit - Replace exact string occurrences (single match only)
find - Search for files using patterns
exec - Execute shell commands with configurable timeouts

These tools enable your agent to work directly with your codebase, configuration files, and project structure.

Full Platform Integration

Agent mode runs with complete ChatBotKit platform capabilities, meaning your CLI agent can:

Access all configured integrations - Slack, Discord, Jira, Linear, Google Workspace, Notion, and 30+ more services
Query datasets - Retrieve information from your knowledge bases
Use skillsets and abilities - Leverage pre-built capabilities and custom functions
Maintain authenticated sessions - No need to manage OAuth flows or API keys for integrated services

This combination of local and remote capabilities is what makes the CLI agent truly powerful.

Simple Command-Line Interface

Getting started is incredibly simple:

# Install globally
npm install --global @chatbotkit/cli

# Set your API token (one-time setup)
export CHATBOTKIT_API_TOKEN=your_token_here

# Run an agent task
cbk agent -p "Create a new file and write hello world to it"

Real-World Use Cases

Development & Coding Assistance

Automated Code Reviews:

cbk agent -p "Read all .js files in src/, check for console.log statements, create a report.md listing the files and line numbers where they appear"

Project Setup:

cbk agent -p "Create a standard Node.js project structure with package.json, src/ folder, README.md, and .gitignore. Include common scripts for testing and building."

Refactoring Helper:

cbk agent -p "Find all files using the old API pattern 'fetchData()', show me the occurrences, and update them to use the new 'getData()' pattern"

Business Process Automation

Document Processing:

cbk agent -p "Read all PDF invoices in the ./invoices folder, extract key data,
create a spreadsheet summary, and send it to #finance Slack channel"

Report Generation:

cbk agent -p "Analyze the last 30 days of application logs, identify error
patterns, create a summary report, and post it to our Notion workspace"

Data Synchronization:

cbk agent -p "Read customer data from local CSV, validate email addresses, cross-reference with our CRM via API, and update records that have changed"

DevOps & System Administration

Deployment Checks:

cbk agent -p "Check git status, verify all tests pass, ensure no console statements in production code, then create a deployment checklist in Jira"

System Monitoring:

cbk agent -p "Check disk usage, analyze recent error logs, if any issues found create alerts in PagerDuty and notify the team via Slack"

Backup Verification:

cbk agent -p "Verify all backup files exist, check their integrity, compare sizes with previous backups, and log results to Google Sheets"

Content Management

Bulk Content Updates:

cbk agent -p "Read all markdown files in content/blog/, update the front matter to include new SEO fields, and create a summary of changes"

Asset Organization:

cbk agent -p "Find all images in assets/, resize images larger than 2MB, optimize them, and update all references in markdown files"

Advanced Configuration

Custom Tool Selection

Choose specific tools for enhanced security or focused functionality:

# Only allow read and write operations
cbk agent -t read,write -p "Create a backup of package.json"

# All tools except exec for safety
cbk agent -t read,write,edit,find -p "Refactor component files"

Model Selection

Use different AI models based on your task requirements:

# Use GPT-5 for complex reasoning
cbk agent -m gpt-5 -p "Analyze codebase and suggest architectural improvements"

# Use Claude for creative tasks
cbk agent -m claude-4.5 -p "Generate comprehensive documentation for all API endpoints"

Environment Configuration

The CLI supports flexible environment variable configuration:

.env.local - Project-specific settings (current directory)
.env - Project configuration (current directory)
~/.cbk/env - Global user configuration (home directory)

Store your API token globally for convenience:

# Create global config file
mkdir -p ~/.cbk
echo "CHATBOTKIT_API_TOKEN=your_token_here" > ~/.cbk/env

Override per project as needed with local .env files.

How It Works

The agent mode leverages the ChatBotKit Agent SDK under the hood, providing:

Autonomous task execution - The agent plans, executes, and adapts without constant supervision
Built-in error handling - Failed operations are communicated to the agent, which can adjust strategy
Progress tracking - System tools like plan, progress, and exit enable sophisticated task management
Type-safe operations - All tools use Zod schemas for validation and error prevention

When you run a command, the agent:

Analyzes your prompt to understand the task
Creates an execution plan
Executes local tools (file operations, shell commands)
Leverages remote integrations when needed (Slack notifications, Jira tickets, etc.)
Reports progress and handles errors gracefully
Exits with a status code when complete

Security Considerations

When using agent mode with local file system and command access:

Review prompts carefully - Ensure your instructions are clear and won't cause unintended side effects
Use tool restrictions - Limit available tools with the t flag when appropriate
Test in safe environments - Try new workflows on test projects first
Monitor execution - The CLI provides real-time feedback on all operations
Backup important data - Especially when testing file modification tasks

The agent only has access to files and commands within your current working directory and its subdirectories.

Getting Started

Installation

npm install --global @chatbotkit/cli@latest

Quick Start

# Set up authentication
export CHATBOTKIT_API_TOKEN=your_token_here

# Try a simple task
cbk agent -p "List all JavaScript files in the current directory"

# More complex automation
cbk agent -p "Read package.json, check for outdated dependencies, create an update plan in updates.md"

Tips for Effective Prompts

Be specific - Clear instructions yield better results
Break down complex tasks - Agent mode handles multi-step workflows well
Mention file paths - Help the agent locate files quickly
Specify output formats - Tell the agent how you want results formatted
Include validation steps - Ask the agent to verify its work

API Management Commands

Beyond agent mode, the CLI also provides direct API access for managing ChatBotKit resources:

# List conversations
cbk api conversation list

# List bots
cbk api bot list

# Manage datasets
cbk api dataset list

Use cbk --help to see all available commands.

What Makes This Unique?

The ChatBotKit CLI with agent mode stands out because it:

Combines local and remote capabilities - No other tool seamlessly bridges local file operations with 30+ cloud integrations
Requires minimal setup - Install, add your token, and start automating
Leverages your existing bots - Use the same ChatBotKit resources you've already configured
Provides autonomous execution - Agents plan and adapt without requiring step-by-step guidance
Maintains security - Scoped access and configurable tool restrictions keep you safe

It's like having a highly capable assistant that understands both your local environment and your cloud infrastructure.

Documentation

For comprehensive documentation including command references, examples, and best practices:

What's Next

We're continuously improving the CLI with upcoming features:

Custom tool plugins - Add your own tools for specialized workflows
Interactive mode - Multi-turn conversations with the agent
Session persistence - Resume long-running tasks across CLI sessions
Workflow templates - Pre-built automation patterns for common scenarios
Enhanced logging - Detailed execution traces for debugging

Examples to Try

Here are some practical examples to get you started:

# Code quality check
cbk agent -p "Check all files for TODO comments and create a todos.md list"

# Documentation generator
cbk agent -p "Read all .js files, extract JSDoc comments, generate API docs in docs/"

# Environment validator
cbk agent -p "Check if all required environment variables from .env.example exist"

# Git workflow
cbk agent -p "Create a feature branch, list uncommitted changes, suggest commit message"

# Project statistics
cbk agent -p "Count lines of code by file type, create a stats report with breakdown"

Join the Community

We'd love to hear about your automation workflows! Share your CLI agent use cases, ask questions, and provide feedback: