14｜Chapter 11: Using Advanced Claude Features — Crazyrouter Series 14

14｜Chapter 11: Using Advanced Claude Features: Crazyrouter Series 14#

This is article 14 in the Crazyrouter Claude Code series. This article focuses on “Chapter 11: Using Advanced Claude Features: Crazyrouter Series 14,” covering Chapter 11: Using Advanced Claude Features, 11.1 Using Custom Tools, and What Custom Tools Are.

Unified access convention: for Claude Code / Anthropic native clients, use ANTHROPIC_BASE_URL=https://cn.crazyrouter.com; for OpenAI-compatible SDKs, HTTP requests, and frontend/backend applications, use base_url=https://cn.crazyrouter.com/v1.

What This Article Covers#

Who it’s for: developers who are using Claude Code, preparing to connect to domestic models, or want to unify their team’s calls through Crazyrouter.
What you’ll learn: how to configure environment variables according to the Crazyrouter documentation, organize workflows, and avoid the /v1/v1/... issue caused by an incorrect Base URL.
Recommended preparation: first create a separate API Token in the Crazyrouter console, then follow the Claude Code integration documentation to complete the basic setup.

Chapter 11: Using Advanced Claude Features#

11.1 Using Custom Tools#

Claude Code supports custom tools, allowing you to extend its capabilities for specific needs.

What Are Custom Tools?#

Custom tools are specific functional modules created by users based on their own needs. They can extend Claude Code’s capabilities.

Key Features of Custom Tools#

Targeted: designed for specific needs
Reusable: can be used multiple times
Configurable: parameters can be adjusted as needed
Integratable: can be integrated with other tools

Types of Custom Tools#

Data processing tools: process data in specific formats
Content generation tools: generate specific types of content
Analysis tools: perform specific analyses
Automation tools: automate specific workflows

How to Find and Use Existing Tools#

Ways to Find Tools#

Official documentation: check the official Claude Code documentation
Community sharing: look for tools shared by users in the community
Plugin marketplace: search for tools in plugin marketplaces
Search engines: use search engines to find tools

Steps to Use a Tool#

Install the tool: install it according to the tool’s instructions
Configure the tool: configure parameters as needed
Use the tool: call the tool in Claude Code
Evaluate the results: assess the tool’s effectiveness
Optimize and adjust: optimize and adjust as needed

How to Configure Simple Tools#

Configuration Example: Data Cleaning Tool#

Requirement: you need a tool to clean Excel data.

Configuration steps:

Define the functionality
- Remove duplicate rows
- Fill blank cells
- Standardize date formats
- Standardize number formats
Create a prompt template

bash

Please help me clean this Excel data:
- Remove duplicate rows
- Fill blank cells with "N/A"
- Standardize the date format as YYYY-MM-DD
- Standardize numbers to two decimal places

Save it as a tool
- Save the prompt template as a tool
- Name it "Data Cleaning Tool"
- Set a quick invocation method
Use the tool
- Upload the Excel file
- Call the "Data Cleaning Tool"
- Get the cleaned data

Configuration Example: Copywriting Generation Tool#

Requirement: you need a tool to generate product copy.

Configuration steps:

Define the functionality
- Generate product descriptions
- Generate ad copy
- Generate social media content
- Generate email content
Create a prompt template

bash

Please help me generate [type] copy:
- Product name: [product name]
- Key features: [feature 1], [feature 2], [feature 3]
- Target users: [target users]
- Style requirements: [style]
- Word count requirement: [word count]

Save it as a tool
- Save the prompt template as a tool
- Name it "Copywriting Generation Tool"
- Set a quick invocation method
Use the tool
- Provide product information
- Call the "Copywriting Generation Tool"
- Get the generated copy

Case Example#

Case: Create a Meeting Minutes Tool#

Requirement: you need a tool to automatically generate meeting minutes.

Configuration steps:

Define the functionality
- Extract key meeting points
- Organize decisions
- List action items
- Generate meeting minutes
Create a prompt template

Markdown

Please help me generate meeting minutes based on the meeting notes:
- Meeting time: [time]
- Meeting location: [location]
- Attendees: [people]
- Meeting content: [content]

Meeting minutes requirements:
- Extract key meeting points
- Organize decisions
- List action items, including owners and deadlines
- Include the schedule for the next meeting

Save it as a tool
- Save the prompt template as a tool
- Name it "Meeting Minutes Tool"
- Set a quick invocation method
Use the tool
- Provide the meeting notes
- Call the "Meeting Minutes Tool"
- Get the generated meeting minutes

Results:

Previous time required: 2–3 hours
Current time required: 30–45 minutes
Time saved: 70–80%

Tips#

Start simple: begin with simple tools and gradually add complexity
Document configuration: record the tool configuration process for future use
Test and validate: after configuration, test and validate the tool’s results
Keep optimizing: continuously optimize the tool based on usage
Share and learn: share useful tools with colleagues and learn from each other

Now, try creating your own custom tool!

11.2 Creating Intelligent Assistants#

Intelligent assistants are an advanced Claude Code feature. You can create dedicated AI assistants that focus on specific domains and provide more professional services.

What Is an Intelligent Assistant?#

An intelligent assistant is an AI assistant configured for specific needs. It has specialized domain knowledge and personalized service capabilities.

Key Features of Intelligent Assistants#

Specialized: focused on a specific domain
Personalized: configured according to individual needs
Efficient: provides more accurate service
Customizable: can be adjusted as needed

Use Cases for Intelligent Assistants#

Professional consulting: provide consulting in professional domains
Content creation: create specific types of content
Data analysis: perform specific types of analysis
Workflow automation: automate specific workflows

How to Create a Dedicated Assistant#

Creation Steps#

Step 1: Define the assistant’s positioning

Define the assistant’s professional domain
Define the assistant’s main functions
Define the assistant’s target users

Step 2: Configure assistant parameters

Set the assistant’s name
Set the assistant’s description
Set the assistant’s professional domain
Set the assistant’s response style

Step 3: Train the assistant’s knowledge

Provide domain knowledge
Provide case examples
Provide best practices

Step 4: Test the assistant’s performance

Test the assistant’s professional capabilities
Test the assistant’s response quality
Test the assistant’s response speed

Step 5: Optimize the assistant configuration

Optimize based on test results
Adjust assistant parameters
Add to the assistant’s knowledge

Creation Example: Marketing Assistant#

Positioning: an intelligent assistant focused on the marketing domain.

Configuration:

Name: Marketing Assistant
Description: an intelligent assistant focused on the marketing domain, providing services such as marketing strategy, copywriting, and data analysis
Professional domain: marketing
Response style: professional, innovative, practical

Knowledge training:

Marketing theory knowledge
Marketing case analysis
Marketing best practices
Marketing tool usage

Testing:

Test marketing strategy suggestions
Test copywriting capabilities
Test data analysis capabilities
Test question-answering capabilities

Configuring the Assistant’s Professional Domain#

Key Points for Professional Domain Configuration#

Knowledge scope: clearly define the assistant’s knowledge scope
Professional depth: determine the assistant’s level of expertise
Update frequency: set the knowledge update frequency
Quality standards: set response quality standards

Configuration Examples#

Domain 1: Marketing domain

Knowledge scope: marketing theory, marketing strategies, marketing tools
Professional depth: intermediate
Update frequency: weekly
Quality standards: professional, practical, innovative

Domain 2: Finance domain

Knowledge scope: financial theory, financial analysis, financial tools
Professional depth: advanced
Update frequency: monthly
Quality standards: accurate, detailed, professional

Domain 3: Human resources domain

Knowledge scope: human resources theory, human resources policies, human resources tools
Professional depth: intermediate
Update frequency: monthly
Quality standards: professional, compliant, practical

Testing the Assistant’s Performance#

Testing Methods#

Functional testing: test the assistant’s main functions
Quality testing: test the assistant’s response quality
Performance testing: test the assistant’s response speed
User experience testing: test the user experience

Test Criteria#

Accuracy: whether the response is accurate
Completeness: whether the response is complete
Practicality: whether the response is practical
Professionalism: whether the response is professional

Test Examples#

Test 1: Marketing strategy recommendations

Question: How can we increase the product's market share?
Expected: Provide specific marketing strategy recommendations
Evaluation: accuracy, completeness, practicality, professionalism

Test 2: Copywriting

Question: Help me write product ad copy
Expected: Provide high-quality advertising copy
Evaluation: accuracy, completeness, practicality, professionalism

Test 3: Data analysis

Question: Help me analyze this sales data
Expected: Provide in-depth data analysis
Evaluation: accuracy, completeness, practicality, professionalism

Case Example#

Case: Creating a Customer Support Assistant#

Positioning: an intelligent assistant focused on customer service

Configuration:

Name: Customer Support Assistant
Description: An intelligent assistant focused on customer service, providing services such as customer inquiries, Q&A, and service recommendations
Domain expertise: customer service
Response style: friendly, professional, patient

Knowledge training:

Customer service theory
Customer service skills
FAQ handling
Customer complaint handling

Testing:

Test customer inquiry handling
Test Q&A capability
Test service recommendation capability
Test complaint handling capability

Results:

Response time: reduced from an average of 4 hours to 30 minutes
Customer satisfaction: increased from 75% to 90%
Improvement: significant

Tips#

Clarify requirements: define the assistant's requirements and positioning clearly
Train continuously: keep training the assistant's knowledge base
Test regularly: regularly test the assistant's performance
Collect feedback: gather user feedback and keep optimizing
Share and learn: share assistant configurations with colleagues and learn from each other

Now, try creating your own intelligent assistant!

11.3 Designing Automated Workflows#

Automated workflows are an important way to improve work efficiency. By designing automated workflows, you can let Claude Code handle repetitive tasks automatically, saving a lot of time and effort.

Identifying Tasks That Can Be Automated#

Characteristics of Automatable Tasks#

Repetitive: tasks that need to be performed repeatedly
Standardized: tasks with clear standards and processes
Rule-based: tasks with explicit rules
Data-driven: tasks based on data processing

Common Automatable Tasks#

Document processing
- Standardizing document formats
- Extracting document content
- Generating documents in bulk
Data processing
- Data cleaning
- Data analysis
- Report generation
Content creation
- Copy generation
- Report writing
- Email replies
Process execution
- Organizing meeting minutes
- Managing to-do items
- Sending notifications

Designing Automated Workflows#

Workflow Design Principles#

Simplicity: the workflow should be simple and easy to understand
Reliability: the workflow should be stable and reliable
Maintainability: the workflow should be easy to maintain
Scalability: the workflow should be easy to extend

Workflow Design Steps#

Step 1: Analyze the task

Identify the task inputs
Identify the task outputs
Identify the task steps

Step 2: Design the workflow

Design the workflow steps
Design the workflow sequence
Design the workflow trigger conditions

Step 3: Implement the workflow

Use Claude Code to implement the workflow
Configure workflow parameters
Test the workflow results

Step 4: Optimize the workflow

Optimize based on test results
Adjust workflow parameters
Improve workflow efficiency

Workflow Design Example#

Task: Automatically generate a weekly report

Step 1: Analyze the task

Input: this week's work data
Output: weekly report document
Steps: collect data, analyze data, generate the report

Step 2: Design the workflow

Export this week's data from the project management tool
Use Claude Code to analyze the data
Use Claude Code to generate the weekly report
Manually review the weekly report
Send the weekly report

Step 3: Implement the workflow

Configure data export
Configure Claude Code analysis
Configure Claude Code generation
Set up manual review
Set up automatic sending

Step 4: Optimize the workflow

Optimize the data export method
Optimize the analysis prompt
Optimize the generation template
Improve review efficiency

Configuring Automation Conditions#

Types of Trigger Conditions#

Time trigger: triggered at a scheduled time interval
Event trigger: triggered by a specific event
Data trigger: triggered by data changes
Manual trigger: triggered manually

Trigger Condition Configuration#

Time trigger configuration

Daily trigger: at a fixed time every day
Weekly trigger: at a fixed time every week
Monthly trigger: at a fixed time every month

Event trigger configuration

File upload: triggered after a file is uploaded
Email received: triggered after an email is received
Task completed: triggered after a task is completed

Data trigger configuration

Data update: triggered after data is updated
Data reaches a threshold: triggered after data reaches a threshold
Data anomaly: triggered when data is abnormal

Configuration Examples#

Configuration 1: Generate a daily report every day

Trigger condition: every day at 18:00
Workflow: collect today's data, generate the daily report, send the daily report

Configuration 2: Generate a weekly report every week

Trigger condition: every Friday at 17:00
Workflow: collect this week's data, generate the weekly report, send the weekly report

Configuration 3: Automatically reply when an email is received

Trigger condition: a specific email is received
Workflow: analyze the email content, generate a reply, send the reply

Testing Automation Results#

Testing Methods#

Functional testing: test the functionality of the automated workflow
Performance testing: test the performance of the automated workflow
Stability testing: test the stability of the automated workflow
User experience testing: test the user experience

Test Metrics#

Accuracy: whether the automated workflow is accurate
Completeness: whether the automated workflow is complete
Efficiency: whether the automated workflow is efficient
Stability: whether the automated workflow is stable

Test Examples#

Test 1: Daily report generation workflow

Test content: test whether daily report generation is accurate
Test method: compare manually generated reports with automatically generated reports
Test result: the automated workflow is accurate, complete, efficient, and stable

Test 2: Email reply workflow

Test content: test whether email replies are accurate
Test method: compare manual replies with automatic replies
Test result: the automated workflow is accurate, complete, efficient, and stable

Case Example#

Case: Automated Meeting Minutes Workflow#

Task: Automatically generate meeting minutes

Workflow design:

Step 1: Meeting recording

Use recording equipment to record the meeting
Automatically save the recording file

Step 2: Speech to text

Use a speech recognition tool
Convert the recording to text

Step 3: Generate meeting minutes

Use Claude Code to generate meeting minutes
Extract key meeting points
Organize decisions
List to-do items

Step 4: Manual review

Manually review the meeting minutes
Adjust inaccurate content
Add missing information

Step 5: Send meeting minutes

Automatically send them to meeting participants
Save the meeting minutes archive

Trigger condition:

Automatically triggered after the meeting ends

Test results:

Accuracy: 95%
Completeness: 90%
Efficiency: saves 80% of the time
Stability: stable and reliable

Results:

Original time required: 2–3 hours
Current time required: 30–45 minutes
Time saved: 70–80%

Tips#

Start with small workflows: begin with simple workflows, then gradually make them more complex
Optimize continuously: keep optimizing workflows based on actual usage
Document configurations: record workflow configurations for future use
Test and validate: after configuration is complete, test and validate the workflow results
Back up configurations: back up workflow configurations to prevent loss

Now, try designing your own automated workflow!

Previous article: Article 13
Next article: Article 15
Configuration guide: Connecting Claude Code to Crazyrouter
API endpoint notes: Base URL and /v1 usage
One-click configuration script: Crazyrouter Claude Code one-click configuration script

Getting Started with Crazyrouter#

If you're ready to connect Claude Code, Chinese models, or your own applications to Crazyrouter through a unified setup, follow this order:

Go to the Crazyrouter console, create a dedicated API Token, and manage permissions separately by project or team.
For Claude Code, use the root domain: https://cn.crazyrouter.com; for OpenAI-compatible SDKs, use: https://cn.crazyrouter.com/v1.
When you need to automatically check the environment or quickly write configuration, use the Crazyrouter Claude Code one-click configuration script.
If debugging fails, check the console logs first, then verify the API Endpoint notes, with special attention to whether an extra /v1 was added to the Base URL. When you need to evaluate model costs or choose a different model, first check the Crazyrouter pricing and models page, then add the models you commonly use to the Token whitelist.

14｜Chapter 11: Using Advanced Claude Features — Crazyrouter Series 14