Claude Agent SDK for Ruby
Disclaimer: This is an unofficial, community-maintained Ruby SDK for Claude Agent. It is not officially supported by Anthropic. For official SDK support, see the Python SDK.
This implementation is based on the official Python SDK and aims to provide feature parity for Ruby developers. Use at your own risk.
Feature Parity with Python SDK (v0.1.48) + Ruby Extras
| Feature | Python | Ruby |
|---|---|---|
One-shot query()
|
query() |
query() |
Bidirectional Client
|
ClaudeSDKClient |
Client |
| Streaming input | AsyncIterable |
Enumerator |
| Custom tools (SDK MCP servers) |
@tool decorator |
create_tool block |
| MCP resources & prompts | ✅ | ✅ |
| Hooks (all 10 events) | ✅ | ✅ |
Permission callbacks (can_use_tool) |
✅ | ✅ |
| Structured output | ✅ | ✅ |
| Thinking config (adaptive/enabled/disabled) | ✅ | ✅ |
| Effort levels | ✅ | ✅ |
| Programmatic subagents | ✅ | ✅ |
| Sandbox settings | ✅ | ✅ |
| Beta features (1M context) | ✅ | ✅ |
| File checkpointing & rewind | ✅ | ✅ |
Session browsing (list_sessions, get_session_messages) |
✅ | ✅ |
Session mutations (rename_session, tag_session) |
✅ | ✅ |
| Task message types (started/progress/notification) | ✅ | ✅ |
| MCP server control (reconnect/toggle/stop) | ✅ | ✅ |
| Subagent context on hook inputs | ✅ | ✅ |
| Typed MCP status response | ✅ | ✅ |
stop_reason on ResultMessage
|
✅ | ✅ |
usage on AssistantMessage
|
✅ | ✅ |
| Fallback model | ✅ | ✅ |
| Plugin support | ✅ | ✅ |
| Custom transport (pluggable I/O layer) | — | ✅ |
| Rails integration (configure block, ActionCable) | — | ✅ |
| Bundled CLI binary | ✅ | — |
Async model
Python uses async/await with anyio (works with both asyncio and Trio). Ruby uses the async gem with fibers — no await keyword needed, blocking calls yield automatically.
# Python
async with ClaudeSDKClient(options) as client:
await client.query("Hello")
async for msg in client.receive_messages():
print(msg)# Ruby
Async do
client = ClaudeAgentSDK::Client.new(options: options)
client.connect
client.query("Hello")
client.receive_messages { |msg| puts msg }
client.disconnect
end.waitCustom tools
Python uses a @tool decorator. Ruby uses create_tool with a block.
# Python
@tool(name="add", description="Add numbers", input_schema={...})
def add(a: int, b: int) -> str:
return str(a + b)# Ruby
add = ClaudeAgentSDK.create_tool("add", "Add numbers", { a: :number, b: :number }) do |args|
{ content: [{ type: "text", text: (args[:a] + args[:b]).to_s }] }
endStreaming input
Python uses AsyncIterable. Ruby uses Enumerator or any #each-able.
# Python
async def messages():
yield {"type": "user", "message": {"role": "user", "content": "Hello"}}
async for msg in query(prompt=messages(), options=options):
print(msg)# Ruby
messages = ClaudeAgentSDK::Streaming.from_array(["Hello", "Follow up"])
ClaudeAgentSDK.query(prompt: messages) { |msg| puts msg }Types
Python uses dataclass with type annotations and TypedDict. Ruby uses plain classes with attr_accessor and keyword args — no runtime type checking, but the same structure.
Configuration defaults
Python passes options directly. Ruby adds ClaudeAgentSDK.configure for global defaults that merge with per-call options — handy for Rails initializers.
# Ruby-only: global defaults
ClaudeAgentSDK.configure do |config|
config.default_options = { model: "sonnet", permission_mode: "bypassPermissions" }
endSubprocess transport
Both SDKs spawn claude CLI as a subprocess with stream-JSON over stdin/stdout. Python uses anyio.open_process; Ruby uses Open3.popen3 with a background Thread for stderr. The wire protocol is identical.
Table of Contents
- Installation
- Quick Start
- Basic Usage: query()
- Client
- Custom Transport
- Custom Tools (SDK MCP Servers)
- Hooks
- Permission Callbacks
- Structured Output
- Thinking Configuration
- Budget Control
- Fallback Model
- Beta Features
- Tools Configuration
- Sandbox Settings
- File Checkpointing & Rewind
- Session Browsing
- Session Mutations
- Rails Integration
- Types
- Error Handling
- Examples
- Development
- License
Installation
Add this line to your application's Gemfile:
# Recommended: Use the latest from GitHub for newest features
gem 'claude-agent-sdk', github: 'ya-luotao/claude-agent-sdk-ruby'
# Or use a stable version from RubyGems
gem 'claude-agent-sdk', '~> 0.11.0'And then execute:
bundle installOr install directly from RubyGems:
gem install claude-agent-sdkPrerequisites:
- Ruby 3.2+
- Node.js
- Claude Code 2.0.0+:
npm install -g @anthropic-ai/claude-code
Agentic Coding Skill
If you're using Claude Code or another agentic coding tool that supports skills, you can install the SDK skill:
npx skills add https://github.com/ya-luotao/claude-agent-sdk-ruby --skill claude-agent-sdk-rubyThis skill teaches your AI coding assistant about the SDK's APIs, patterns, and best practices, making it easier to get help writing code that uses this SDK.
Quick Start
require 'claude_agent_sdk'
ClaudeAgentSDK.query(prompt: "What is 2 + 2?") do |message|
puts message
endBasic Usage: query()
query() is a function for querying Claude Code. It yields response messages to a block.
require 'claude_agent_sdk'
# Simple query
ClaudeAgentSDK.query(prompt: "Hello Claude") do |message|
if message.is_a?(ClaudeAgentSDK::AssistantMessage)
message.content.each do |block|
puts block.text if block.is_a?(ClaudeAgentSDK::TextBlock)
end
end
end
# With options
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
system_prompt: "You are a helpful assistant",
max_turns: 1
)
ClaudeAgentSDK.query(prompt: "Tell me a joke", options: options) do |message|
puts message
endUsing Tools
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
allowed_tools: ['Read', 'Write', 'Bash'],
permission_mode: 'acceptEdits' # auto-accept file edits
)
ClaudeAgentSDK.query(
prompt: "Create a hello.rb file",
options: options
) do |message|
# Process tool use and results
endWorking Directory
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
cwd: "/path/to/project"
)Streaming Input
The query() function supports streaming input, allowing you to send multiple messages dynamically instead of a single prompt string.
require 'claude_agent_sdk'
# Create a stream of messages
messages = ['Hello!', 'What is 2+2?', 'Thanks!']
stream = ClaudeAgentSDK::Streaming.from_array(messages)
# Query with streaming input
ClaudeAgentSDK.query(prompt: stream) do |message|
puts message if message.is_a?(ClaudeAgentSDK::AssistantMessage)
endYou can also create custom streaming enumerators:
# Dynamic message generation
stream = Enumerator.new do |yielder|
yielder << ClaudeAgentSDK::Streaming.user_message("First message")
# Do some processing...
yielder << ClaudeAgentSDK::Streaming.user_message("Second message")
yielder << ClaudeAgentSDK::Streaming.user_message("Third message")
end
ClaudeAgentSDK.query(prompt: stream) do |message|
# Process responses
endFor a complete example, see examples/streaming_input_example.rb.
Client
ClaudeAgentSDK::Client supports bidirectional, interactive conversations with Claude Code. Unlike query(), Client enables custom tools, hooks, and permission callbacks, all of which can be defined as Ruby procs/lambdas.
The Client class automatically uses streaming mode for bidirectional communication, allowing you to send multiple queries dynamically during a single session without closing the connection.
Basic Client Usage
require 'claude_agent_sdk'
require 'async'
Async do
client = ClaudeAgentSDK::Client.new
begin
# Connect automatically uses streaming mode for bidirectional communication
client.connect
# Send a query
client.query("What is the capital of France?")
# Receive the response
client.receive_response do |msg|
if msg.is_a?(ClaudeAgentSDK::AssistantMessage)
msg.content.each do |block|
puts block.text if block.is_a?(ClaudeAgentSDK::TextBlock)
end
elsif msg.is_a?(ClaudeAgentSDK::ResultMessage)
puts "Cost: $#{msg.total_cost_usd}" if msg.total_cost_usd
end
end
ensure
client.disconnect
end
end.waitAdvanced Client Features
Async do
client = ClaudeAgentSDK::Client.new
client.connect
# Send interrupt signal
client.interrupt
# Change permission mode during conversation
client.set_permission_mode('acceptEdits')
# Change AI model during conversation
client.set_model('claude-sonnet-4-5')
# Get MCP server connection status
status = client.get_mcp_status
puts "MCP status: #{status}"
# Get server initialization info
info = client.get_server_info
puts "Available commands: #{info}"
# Reconnect a failed MCP server
client.reconnect_mcp_server('my-server')
# Enable or disable an MCP server
client.toggle_mcp_server('my-server', false)
# Stop a running background task
client.stop_task('task_abc123')
client.disconnect
end.waitCustom Transport
By default, Client uses SubprocessCLITransport to spawn the Claude Code CLI locally. You can provide a custom transport class to connect via other channels (e.g., E2B sandbox, remote SSH, WebSocket):
# Custom transport must implement the Transport interface:
# connect, write, read_messages, end_input, close, ready?
class E2BSandboxTransport < ClaudeAgentSDK::Transport
def initialize(options, sandbox:)
@options = options
@sandbox = sandbox
end
def connect
@sandbox.connect
end
def write(data)
@sandbox.stdin_write(data)
end
def read_messages(&block)
@sandbox.stdout_read_lines { |line| yield JSON.parse(line, symbolize_names: true) }
end
def end_input
@sandbox.close_stdin
end
def close
@sandbox.disconnect
end
def ready?
@sandbox.connected?
end
end
# Use it with Client — all connect orchestration (option transforms,
# MCP extraction, hook conversion, Query lifecycle) is handled for you
Async do
client = ClaudeAgentSDK::Client.new(
options: options,
transport_class: E2BSandboxTransport,
transport_args: { sandbox: my_sandbox }
)
client.connect
client.query("Hello from the sandbox!")
client.receive_response { |msg| puts msg }
client.disconnect
end.waitCustom Tools (SDK MCP Servers)
A custom tool is a Ruby proc/lambda that you can offer to Claude, for Claude to invoke as needed.
Custom tools are implemented as in-process MCP servers that run directly within your Ruby application, eliminating the need for separate processes that regular MCP servers require.
Implementation: This SDK uses the official Ruby MCP SDK (mcp gem) internally, providing full protocol compliance while offering a simpler block-based API for tool definition.
Creating a Simple Tool
require 'claude_agent_sdk'
require 'async'
# Define a tool using create_tool (with optional annotations)
greet_tool = ClaudeAgentSDK.create_tool(
'greet', 'Greet a user', { name: :string },
annotations: { title: 'Greeter', readOnlyHint: true }
) do |args|
{ content: [{ type: 'text', text: "Hello, #{args[:name]}!" }] }
end
# Create an SDK MCP server
server = ClaudeAgentSDK.create_sdk_mcp_server(
name: 'my-tools',
version: '1.0.0',
tools: [greet_tool]
)
# Use it with Claude
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
mcp_servers: { tools: server },
allowed_tools: ['mcp__tools__greet']
)
Async do
client = ClaudeAgentSDK::Client.new(options: options)
client.connect
client.query("Greet Alice")
client.receive_response { |msg| puts msg }
client.disconnect
end.waitPre-built JSON Schemas
If your schemas come from another library (e.g., RubyLLM) that deep-stringifies keys, the SDK handles them transparently — both symbol-keyed and string-keyed schemas are accepted and normalized:
# Symbol keys (standard Ruby)
tool = ClaudeAgentSDK.create_tool('save', 'Save a fact', {
type: 'object',
properties: { fact: { type: 'string' } },
required: ['fact']
}) { |args| { content: [{ type: 'text', text: "Saved: #{args[:fact]}" }] } }
# String keys (e.g., from RubyLLM or JSON.parse)
tool = ClaudeAgentSDK.create_tool('save', 'Save a fact', {
'type' => 'object',
'properties' => { 'fact' => { 'type' => 'string' } },
'required' => ['fact']
}) { |args| { content: [{ type: 'text', text: "Saved: #{args[:fact]}" }] } }Benefits Over External MCP Servers
- No subprocess management - Runs in the same process as your application
- Better performance - No IPC overhead for tool calls
- Simpler deployment - Single Ruby process instead of multiple
- Easier debugging - All code runs in the same process
- Direct access - Tools can directly access your application's state
Calculator Example
# Define calculator tools
add_tool = ClaudeAgentSDK.create_tool('add', 'Add two numbers', { a: :number, b: :number }) do |args|
result = args[:a] + args[:b]
{ content: [{ type: 'text', text: "#{args[:a]} + #{args[:b]} = #{result}" }] }
end
divide_tool = ClaudeAgentSDK.create_tool('divide', 'Divide numbers', { a: :number, b: :number }) do |args|
if args[:b] == 0
{ content: [{ type: 'text', text: 'Error: Division by zero' }], is_error: true }
else
result = args[:a] / args[:b]
{ content: [{ type: 'text', text: "Result: #{result}" }] }
end
end
# Create server
calculator = ClaudeAgentSDK.create_sdk_mcp_server(
name: 'calculator',
tools: [add_tool, divide_tool]
)
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
mcp_servers: { calc: calculator },
allowed_tools: ['mcp__calc__add', 'mcp__calc__divide']
)Mixed Server Support
You can use both SDK and external MCP servers together:
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
mcp_servers: {
internal: sdk_server, # In-process SDK server
external: { # External subprocess server
type: 'stdio',
command: 'external-server'
}
}
)MCP Resources and Prompts
SDK MCP servers can also expose resources (data sources) and prompts (reusable templates):
# Create a resource (data source Claude can read)
config_resource = ClaudeAgentSDK.create_resource(
uri: 'config://app/settings',
name: 'Application Settings',
description: 'Current app configuration',
mime_type: 'application/json'
) do
config_data = { app_name: 'MyApp', version: '1.0.0' }
{
contents: [{
uri: 'config://app/settings',
mimeType: 'application/json',
text: JSON.pretty_generate(config_data)
}]
}
end
# Create a prompt template
review_prompt = ClaudeAgentSDK.create_prompt(
name: 'code_review',
description: 'Review code for best practices',
arguments: [
{ name: 'code', description: 'Code to review', required: true }
]
) do |args|
{
messages: [{
role: 'user',
content: {
type: 'text',
text: "Review this code: #{args[:code]}"
}
}]
}
end
# Create server with tools, resources, and prompts
server = ClaudeAgentSDK.create_sdk_mcp_server(
name: 'dev-tools',
tools: [my_tool],
resources: [config_resource],
prompts: [review_prompt]
)For complete examples, see examples/mcp_calculator.rb and examples/mcp_resources_prompts_example.rb.
Hooks
A hook is a Ruby proc/lambda that the Claude Code application (not Claude) invokes at specific points of the Claude agent loop. Hooks can provide deterministic processing and automated feedback for Claude. Read more in Claude Code Hooks Reference.
Supported Events
All hook input objects include common fields like session_id, transcript_path, cwd, and permission_mode.
-
PreToolUse→PreToolUseHookInput(tool_name,tool_input,tool_use_id) -
PostToolUse→PostToolUseHookInput(tool_name,tool_input,tool_response,tool_use_id) -
PostToolUseFailure→PostToolUseFailureHookInput(tool_name,tool_input,tool_use_id,error,is_interrupt) -
UserPromptSubmit→UserPromptSubmitHookInput(prompt) -
Stop→StopHookInput(stop_hook_active) -
SubagentStop→SubagentStopHookInput(stop_hook_active,agent_id,agent_transcript_path,agent_type) -
PreCompact→PreCompactHookInput(trigger,custom_instructions) -
Notification→NotificationHookInput(message,title,notification_type) -
SubagentStart→SubagentStartHookInput(agent_id,agent_type) -
PermissionRequest→PermissionRequestHookInput(tool_name,tool_input,permission_suggestions)
Example
require 'claude_agent_sdk'
require 'async'
Async do
# Define a hook that blocks dangerous bash commands
bash_hook = lambda do |input, _tool_use_id, _context|
# Hook inputs are typed objects (e.g., PreToolUseHookInput) with Ruby-style accessors
return {} unless input.respond_to?(:tool_name) && input.tool_name == 'Bash'
tool_input = input.tool_input || {}
command = tool_input[:command] || tool_input['command'] || ''
block_patterns = ['rm -rf', 'foo.sh']
block_patterns.each do |pattern|
if command.include?(pattern)
return {
hookSpecificOutput: {
hookEventName: 'PreToolUse',
permissionDecision: 'deny',
permissionDecisionReason: "Command contains forbidden pattern: #{pattern}"
}
}
end
end
{} # Allow if no patterns match
end
# Create options with hook
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
allowed_tools: ['Bash'],
hooks: {
'PreToolUse' => [
ClaudeAgentSDK::HookMatcher.new(
matcher: 'Bash',
hooks: [bash_hook]
)
]
}
)
client = ClaudeAgentSDK::Client.new(options: options)
client.connect
# Test: Command with forbidden pattern (will be blocked)
client.query("Run the bash command: ./foo.sh --help")
client.receive_response { |msg| puts msg }
client.disconnect
end.waitFor more examples, see examples/hooks_example.rb.
Permission Callbacks
A permission callback is a Ruby proc/lambda that allows you to programmatically control tool execution. This gives you fine-grained control over what tools Claude can use and with what inputs.
Example
require 'claude_agent_sdk'
require 'async'
Async do
# Define a permission callback
permission_callback = lambda do |tool_name, input, context|
# Allow Read operations
if tool_name == 'Read'
return ClaudeAgentSDK::PermissionResultAllow.new
end
# Block Write to sensitive files
if tool_name == 'Write'
file_path = input[:file_path] || input['file_path']
if file_path && file_path.include?('/etc/')
return ClaudeAgentSDK::PermissionResultDeny.new(
message: 'Cannot write to sensitive system files',
interrupt: false
)
end
return ClaudeAgentSDK::PermissionResultAllow.new
end
# Default: allow
ClaudeAgentSDK::PermissionResultAllow.new
end
# Create options with permission callback
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
allowed_tools: ['Read', 'Write', 'Bash'],
can_use_tool: permission_callback
)
client = ClaudeAgentSDK::Client.new(options: options)
client.connect
# This will be allowed
client.query("Create a file called test.txt with content 'Hello'")
client.receive_response { |msg| puts msg }
# This will be blocked
client.query("Write to /etc/passwd")
client.receive_response { |msg| puts msg }
client.disconnect
end.waitFor more examples, see examples/permission_callback_example.rb.
Structured Output
Use output_format to get validated JSON responses matching a schema. The Claude CLI returns structured output via a StructuredOutput tool use block.
require 'claude_agent_sdk'
require 'json'
# Define a JSON schema
schema = {
type: 'object',
properties: {
name: { type: 'string' },
age: { type: 'integer' },
skills: { type: 'array', items: { type: 'string' } }
},
required: %w[name age skills]
}
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
output_format: { type: 'json_schema', schema: schema },
max_turns: 3
)
structured_data = nil
ClaudeAgentSDK.query(
prompt: "Create a profile for a software engineer",
options: options
) do |message|
if message.is_a?(ClaudeAgentSDK::AssistantMessage)
message.content.each do |block|
# Structured output comes via StructuredOutput tool use
if block.is_a?(ClaudeAgentSDK::ToolUseBlock) && block.name == 'StructuredOutput'
structured_data = block.input
end
end
end
end
if structured_data
puts "Name: #{structured_data[:name]}"
puts "Age: #{structured_data[:age]}"
puts "Skills: #{structured_data[:skills].join(', ')}"
endFor complete examples, see examples/structured_output_example.rb.
Thinking Configuration
Control extended thinking behavior with typed configuration objects. The thinking option takes precedence over the deprecated max_thinking_tokens.
# Adaptive thinking — uses a default budget of 32,000 tokens
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
thinking: ClaudeAgentSDK::ThinkingConfigAdaptive.new
)
# Enabled thinking with custom budget
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
thinking: ClaudeAgentSDK::ThinkingConfigEnabled.new(budget_tokens: 50_000)
)
# Explicitly disabled thinking
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
thinking: ClaudeAgentSDK::ThinkingConfigDisabled.new
)Use the effort option to control the model's effort level:
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
effort: 'high' # 'low', 'medium', or 'high'
)Note: When
system_promptisnil(the default), the SDK passes--system-prompt ""to the CLI, which suppresses the default Claude Code system prompt. To use the default system prompt, use aSystemPromptPreset.
Budget Control
Use max_budget_usd to set a spending cap for your queries:
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
max_budget_usd: 0.10, # Cap at $0.10
max_turns: 3
)
ClaudeAgentSDK.query(prompt: "Explain recursion", options: options) do |message|
if message.is_a?(ClaudeAgentSDK::ResultMessage)
puts "Cost: $#{message.total_cost_usd}"
end
endFor complete examples, see examples/budget_control_example.rb.
Fallback Model
Use fallback_model to specify a backup model if the primary is unavailable:
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
model: 'claude-sonnet-4-20250514',
fallback_model: 'claude-3-5-haiku-20241022'
)
ClaudeAgentSDK.query(prompt: "Hello", options: options) do |message|
if message.is_a?(ClaudeAgentSDK::AssistantMessage)
puts "Model used: #{message.model}"
end
endFor complete examples, see examples/fallback_model_example.rb.
Beta Features
Enable experimental features using the betas option:
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
betas: ['context-1m-2025-08-07'] # Extended context window
)
ClaudeAgentSDK.query(prompt: "Analyze this large document...", options: options) do |message|
puts message
endAvailable beta features are listed in the SDK_BETAS constant.
Tools Configuration
Configure base tools separately from allowed tools:
# Using an array of tool names
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
tools: ['Read', 'Edit', 'Bash'] # Base tools available
)
# Using a preset
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
tools: ClaudeAgentSDK::ToolsPreset.new(preset: 'claude_code')
)
# Appending to allowed tools
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
append_allowed_tools: ['Write', 'Bash']
)Sandbox Settings
Run commands in an isolated sandbox for additional security:
sandbox = ClaudeAgentSDK::SandboxSettings.new(
enabled: true,
auto_allow_bash_if_sandboxed: true,
network: ClaudeAgentSDK::SandboxNetworkConfig.new(
allow_local_binding: true
)
)
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
sandbox: sandbox,
permission_mode: 'acceptEdits'
)
ClaudeAgentSDK.query(prompt: "Run some commands", options: options) do |message|
puts message
endFile Checkpointing & Rewind
Enable file checkpointing to revert file changes to a previous state:
require 'async'
Async do
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
enable_file_checkpointing: true,
permission_mode: 'acceptEdits'
)
client = ClaudeAgentSDK::Client.new(options: options)
client.connect
# Track user message UUIDs for potential rewind
user_message_uuids = []
# First query - create a file
client.query("Create a test.rb file with some code")
client.receive_response do |message|
# Process all message types as needed
case message
when ClaudeAgentSDK::UserMessage
# Capture UUID for rewind capability
user_message_uuids << message.uuid if message.uuid
when ClaudeAgentSDK::AssistantMessage
# Handle assistant responses
message.content.each do |block|
puts block.text if block.is_a?(ClaudeAgentSDK::TextBlock)
end
when ClaudeAgentSDK::ResultMessage
puts "Query completed (cost: $#{message.total_cost_usd})"
end
end
# Second query - modify the file
client.query("Modify the test.rb file to add error handling")
client.receive_response do |message|
user_message_uuids << message.uuid if message.is_a?(ClaudeAgentSDK::UserMessage) && message.uuid
end
# Rewind to the first checkpoint (undoes the second query's changes)
if user_message_uuids.first
puts "Rewinding to checkpoint: #{user_message_uuids.first}"
client.rewind_files(user_message_uuids.first)
end
client.disconnect
end.waitNote: The
uuidfield onUserMessageis populated by the CLI and represents checkpoint identifiers. Rewinding to a UUID restores file state to what it was at that point in the conversation.
Session Browsing
Browse and inspect previous Claude Code sessions directly from Ruby — no CLI subprocess required.
Listing Sessions
# List all sessions (sorted by most recent first)
sessions = ClaudeAgentSDK.list_sessions
sessions.each do |session|
puts "#{session.session_id}: #{session.summary} (#{session.git_branch})"
end
# List sessions for a specific directory
sessions = ClaudeAgentSDK.list_sessions(directory: '/path/to/project', limit: 10)
# Include git worktree sessions
sessions = ClaudeAgentSDK.list_sessions(directory: '.', include_worktrees: true)Each SDKSessionInfo includes:
-
session_id,summary,last_modified,file_size -
custom_title,first_prompt,git_branch,cwd
Reading Session Messages
# Get the full conversation from a session
messages = ClaudeAgentSDK.get_session_messages(session_id: 'abc-123-...')
messages.each do |msg|
puts "[#{msg.type}] #{msg.message}"
end
# Paginate through messages
page = ClaudeAgentSDK.get_session_messages(session_id: 'abc-123-...', offset: 10, limit: 20)Each SessionMessage includes type ("user" or "assistant"), uuid, session_id, and message (raw API dict).
Note: Session browsing reads
~/.claude/projects/JSONL files directly. It respects theCLAUDE_CONFIG_DIRenvironment variable and automatically detects git worktrees.
Session Mutations
Rename or tag sessions programmatically — no CLI subprocess required.
Renaming a Session
# Rename a session (appends a custom-title JSONL entry)
ClaudeAgentSDK.rename_session(
session_id: '550e8400-e29b-41d4-a716-446655440000',
title: 'My refactoring session',
directory: '/path/to/project' # optional
)Tagging a Session
# Tag a session (Unicode-sanitized before storing)
ClaudeAgentSDK.tag_session(
session_id: '550e8400-e29b-41d4-a716-446655440000',
tag: 'experiment'
)
# Clear a tag
ClaudeAgentSDK.tag_session(
session_id: '550e8400-e29b-41d4-a716-446655440000',
tag: nil
)Note: Session mutations use append-only JSONL writes with
O_WRONLY | O_APPEND(noO_CREAT) for TOCTOU safety. They are safe to call while the session is open in a CLI process.
Rails Integration
The SDK integrates well with Rails applications. Here are common patterns:
ActionCable Streaming
Stream Claude responses to the frontend in real-time:
# app/jobs/chat_agent_job.rb
class ChatAgentJob < ApplicationJob
queue_as :claude_agents
def perform(chat_id, message_content)
Async do
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
system_prompt: { type: 'preset', preset: 'claude_code' },
permission_mode: 'bypassPermissions'
)
client = ClaudeAgentSDK::Client.new(options: options)
begin
client.connect
client.query(message_content)
client.receive_response do |message|
case message
when ClaudeAgentSDK::AssistantMessage
text = extract_text(message)
ChatChannel.broadcast_to(chat_id, { type: 'chunk', content: text })
when ClaudeAgentSDK::ResultMessage
ChatChannel.broadcast_to(chat_id, {
type: 'complete',
content: message.result,
cost: message.total_cost_usd
})
end
end
ensure
client.disconnect
end
end.wait
end
private
def extract_text(message)
message.content
.select { |b| b.is_a?(ClaudeAgentSDK::TextBlock) }
.map(&:text)
.join("\n\n")
end
endSession Resumption
Persist Claude sessions for multi-turn conversations:
# app/models/chat_session.rb
class ChatSession < ApplicationRecord
# Columns: id, claude_session_id, user_id, created_at, updated_at
def send_message(content)
options = build_options
client = ClaudeAgentSDK::Client.new(options: options)
Async do
client.connect
client.query(content, session_id: claude_session_id ? nil : generate_session_id)
client.receive_response do |message|
if message.is_a?(ClaudeAgentSDK::ResultMessage)
# Save session ID for next message
update!(claude_session_id: message.session_id)
end
end
ensure
client.disconnect
end.wait
end
private
def build_options
opts = {
permission_mode: 'bypassPermissions',
setting_sources: []
}
opts[:resume] = claude_session_id if claude_session_id.present?
ClaudeAgentSDK::ClaudeAgentOptions.new(**opts)
end
def generate_session_id
"chat_#{id}_#{Time.current.to_i}"
end
endBackground Jobs with Error Handling
class ClaudeAgentJob < ApplicationJob
queue_as :claude_agents
retry_on ClaudeAgentSDK::ProcessError, wait: :polynomially_longer, attempts: 3
def perform(task_id)
task = Task.find(task_id)
Async do
execute_agent(task)
end.wait
rescue ClaudeAgentSDK::CLINotFoundError => e
task.update!(status: 'failed', error: 'Claude CLI not installed')
raise
end
private
def execute_agent(task)
# ... agent execution
end
endHTTP MCP Servers
Connect to remote tool services:
mcp_servers = {
'api_tools' => ClaudeAgentSDK::McpHttpServerConfig.new(
url: ENV['MCP_SERVER_URL'],
headers: { 'Authorization' => "Bearer #{ENV['MCP_TOKEN']}" }
).to_h
}
options = ClaudeAgentSDK::ClaudeAgentOptions.new(
mcp_servers: mcp_servers,
permission_mode: 'bypassPermissions'
)For complete examples, see:
- examples/rails_actioncable_example.rb
- examples/rails_background_job_example.rb
- examples/session_resumption_example.rb
- examples/http_mcp_server_example.rb
Types
See lib/claude_agent_sdk/types.rb for complete type definitions.
Message Types
# Union type of all possible messages
Message = UserMessage | AssistantMessage | SystemMessage | ResultMessageUserMessage
User input message.
class UserMessage
attr_accessor :content, # String | Array<ContentBlock>
:uuid, # String | nil - Unique ID for rewind support
:parent_tool_use_id, # String | nil
:tool_use_result # Hash | nil - Tool result data when message is a tool response
endAssistantMessage
Assistant response message with content blocks.
class AssistantMessage
attr_accessor :content, # Array<ContentBlock>
:model, # String
:parent_tool_use_id,# String | nil
:error, # String | nil ('authentication_failed', 'billing_error', 'rate_limit', 'invalid_request', 'server_error', 'unknown')
:usage # Hash | nil - Token usage info from the API response
endSystemMessage
System message with metadata. Task lifecycle events are typed subclasses.
class SystemMessage
attr_accessor :subtype, # String ('init', 'task_started', 'task_progress', 'task_notification', etc.)
:data # Hash
end
# Typed subclasses (all inherit from SystemMessage, so is_a?(SystemMessage) still works)
class TaskStartedMessage < SystemMessage
attr_accessor :task_id, :description, :uuid, :session_id, :tool_use_id, :task_type
end
class TaskProgressMessage < SystemMessage
attr_accessor :task_id, :description, :usage, :uuid, :session_id, :tool_use_id, :last_tool_name
end
class TaskNotificationMessage < SystemMessage
attr_accessor :task_id, :status, :output_file, :summary, :uuid, :session_id, :tool_use_id, :usage
endResultMessage
Final result message with cost and usage information.
class ResultMessage
attr_accessor :subtype, # String
:duration_ms, # Integer
:duration_api_ms, # Integer
:is_error, # Boolean
:num_turns, # Integer
:session_id, # String
:stop_reason, # String | nil ('end_turn', 'max_tokens', 'stop_sequence')
:total_cost_usd, # Float | nil
:usage, # Hash | nil
:result, # String | nil (final text result)
:structured_output # Hash | nil (when using output_format)
endContent Block Types
# Union type of all content blocks
ContentBlock = TextBlock | ThinkingBlock | ToolUseBlock | ToolResultBlock | UnknownBlockTextBlock
Text content block.
class TextBlock
attr_accessor :text # String
endThinkingBlock
Thinking content block (for models with extended thinking capability).
class ThinkingBlock
attr_accessor :thinking, # String
:signature # String
endToolUseBlock
Tool use request block.
class ToolUseBlock
attr_accessor :id, # String
:name, # String
:input # Hash
endToolResultBlock
Tool execution result block.
class ToolResultBlock
attr_accessor :tool_use_id, # String
:content, # String | Array<Hash> | nil
:is_error # Boolean | nil
endUnknownBlock
Generic content block for types the SDK doesn't explicitly handle (e.g., document for PDFs, image for inline images). Preserves the raw data for forward compatibility with newer CLI versions.
class UnknownBlock
attr_accessor :type, # String — the original block type (e.g., "document")
:data # Hash — the full raw block hash
endError Types
# Base exception class for all SDK errors
class ClaudeSDKError < StandardError; end
# Raised when connection to Claude Code fails
class CLIConnectionError < ClaudeSDKError; end
# Raised when the control protocol does not respond in time
class ControlRequestTimeoutError < CLIConnectionError; end
# Raised when Claude Code CLI is not found
class CLINotFoundError < CLIConnectionError
# @param message [String] Error message (default: "Claude Code not found")
# @param cli_path [String, nil] Optional path to the CLI that was not found
end
# Raised when the Claude Code process fails
class ProcessError < ClaudeSDKError
attr_reader :exit_code, # Integer | nil
:stderr # String | nil
end
# Raised when JSON parsing fails
class CLIJSONDecodeError < ClaudeSDKError
attr_reader :line, # String - The line that failed to parse
:original_error # Exception - The original JSON decode exception
end
# Raised when message parsing fails
class MessageParseError < ClaudeSDKError
attr_reader :data # Hash | nil
endConfiguration Types
| Type | Description |
|---|---|
Configuration |
Global defaults via ClaudeAgentSDK.configure block |
ClaudeAgentOptions |
Main configuration for queries and clients |
HookMatcher |
Hook configuration with matcher pattern and timeout |
PermissionResultAllow |
Permission callback result to allow tool use |
PermissionResultDeny |
Permission callback result to deny tool use |
AgentDefinition |
Agent definition with description, prompt, tools, model, skills, memory, mcp_servers |
ThinkingConfigAdaptive |
Adaptive thinking mode (32,000 token default budget) |
ThinkingConfigEnabled |
Enabled thinking with explicit budget_tokens
|
ThinkingConfigDisabled |
Disabled thinking (0 tokens) |
SdkMcpTool |
SDK MCP tool definition with name, description, input_schema, handler, annotations |
McpStdioServerConfig |
MCP server config for stdio transport |
McpSSEServerConfig |
MCP server config for SSE transport |
McpHttpServerConfig |
MCP server config for HTTP transport |
SdkPluginConfig |
SDK plugin configuration |
McpServerStatus |
Status of a single MCP server connection (with .parse) |
McpStatusResponse |
Response from get_mcp_status containing all server statuses (with .parse) |
McpServerInfo |
MCP server name and version |
McpToolInfo |
MCP tool name, description, and annotations |
McpToolAnnotations |
MCP tool annotation hints (read_only, destructive, open_world) |
TaskUsage |
Typed usage data (total_tokens, tool_uses, duration_ms) with from_hash factory |
SDKSessionInfo |
Session metadata from list_sessions
|
SessionMessage |
Single message from get_session_messages
|
SandboxSettings |
Sandbox settings for isolated command execution |
SandboxNetworkConfig |
Network configuration for sandbox |
SandboxIgnoreViolations |
Configure which sandbox violations to ignore |
SystemPromptPreset |
System prompt preset configuration |
ToolsPreset |
Tools preset configuration for base tools selection |
Constants
| Constant | Description |
|---|---|
SDK_BETAS |
Available beta features (e.g., "context-1m-2025-08-07") |
PERMISSION_MODES |
Available permission modes |
SETTING_SOURCES |
Available setting sources |
HOOK_EVENTS |
Available hook events |
ASSISTANT_MESSAGE_ERRORS |
Possible error types in AssistantMessage |
TASK_NOTIFICATION_STATUSES |
Task lifecycle notification statuses (completed, failed, stopped) |
MCP_SERVER_CONNECTION_STATUSES |
MCP server connection states (connected, failed, needs-auth, pending, disabled) |
Error Handling
AssistantMessage Errors
AssistantMessage includes an error field for API-level errors:
ClaudeAgentSDK.query(prompt: "Hello") do |message|
if message.is_a?(ClaudeAgentSDK::AssistantMessage) && message.error
case message.error
when 'rate_limit'
puts "Rate limited - retry after delay"
when 'authentication_failed'
puts "Check your API key"
when 'billing_error'
puts "Check your billing status"
when 'invalid_request'
puts "Invalid request format"
when 'server_error'
puts "Server error - retry later"
end
end
endFor complete examples, see examples/error_handling_example.rb.
Exception Handling
require 'claude_agent_sdk'
begin
ClaudeAgentSDK.query(prompt: "Hello") do |message|
puts message
end
rescue ClaudeAgentSDK::ControlRequestTimeoutError
puts "Control protocol timed out — consider increasing the timeout"
rescue ClaudeAgentSDK::CLINotFoundError
puts "Please install Claude Code"
rescue ClaudeAgentSDK::ProcessError => e
puts "Process failed with exit code: #{e.exit_code}"
rescue ClaudeAgentSDK::CLIJSONDecodeError => e
puts "Failed to parse response: #{e}"
endConfiguring Timeout
The control request timeout defaults to 1200 seconds (20 minutes) to accommodate long-running agent sessions. Override it via environment variable:
# Set a custom timeout (in seconds)
export CLAUDE_AGENT_SDK_CONTROL_REQUEST_TIMEOUT_SECONDS=300 # 5 minutesError Types
| Error | Description |
|---|---|
ClaudeSDKError |
Base error for all SDK errors |
CLIConnectionError |
Connection issues |
ControlRequestTimeoutError |
Control protocol timeout (configurable via env var) |
CLINotFoundError |
Claude Code not installed |
ProcessError |
Process failed (includes exit_code and stderr) |
CLIJSONDecodeError |
JSON parsing issues |
MessageParseError |
Message parsing issues |
See lib/claude_agent_sdk/errors.rb for all error types.
Available Tools
See the Claude Code documentation for a complete list of available tools.
Examples
Core Examples
| Example | Description |
|---|---|
| examples/quick_start.rb | Basic query() usage with options |
| examples/client_example.rb | Interactive Client usage |
| examples/streaming_input_example.rb | Streaming input for multi-turn conversations |
| examples/session_resumption_example.rb | Multi-turn conversations with session persistence |
| examples/structured_output_example.rb | JSON schema structured output |
| examples/error_handling_example.rb | Error handling with AssistantMessage.error
|
MCP Server Examples
| Example | Description |
|---|---|
| examples/mcp_calculator.rb | Custom tools with SDK MCP servers |
| examples/mcp_resources_prompts_example.rb | MCP resources and prompts |
| examples/http_mcp_server_example.rb | HTTP/SSE MCP server configuration |
Hooks & Permissions
| Example | Description |
|---|---|
| examples/hooks_example.rb | Using hooks to control tool execution |
| examples/advanced_hooks_example.rb | Typed hook inputs/outputs |
| examples/permission_callback_example.rb | Dynamic tool permission control |
Advanced Features
| Example | Description |
|---|---|
| examples/budget_control_example.rb | Budget control with max_budget_usd
|
| examples/fallback_model_example.rb | Fallback model configuration |
| examples/extended_thinking_example.rb | Extended thinking (API parity) |
Rails Integration
| Example | Description |
|---|---|
| examples/rails_actioncable_example.rb | ActionCable streaming to frontend |
| examples/rails_background_job_example.rb | Background jobs with session resumption |
Development
After checking out the repo, run bundle install to install dependencies. Then, run bundle exec rspec to run the tests.
License
The gem is available as open source under the terms of the MIT License.