Claude Projects

From an earlier video

This is material from an earlier video — Claude’s interface and features may have changed since. The idea and the approach still hold.

Projects in Claude: automating routine tasks for professionals

Projects in Claude is a powerful system for automating recurring tasks that works far more effectively than the equivalent features in ChatGPT.

Projects let you create specialised AI assistants with their own knowledge base, system instructions and integrations with external services.

What are projects in Claude?

A project is an isolated workspace with:

• System instructions — how Claude should behave in this project
• A knowledge base — attached files, documents, data
• Integrations — connections to Google Drive, GitHub, Notion
• Conversation history — keeping the context for the project

🔄 How it works in everyday life

The traditional approach:

1. You open a new chat
2. You explain the task’s context
3. You attach the necessary files
4. You give detailed instructions
5. You get the result
6. You repeat the whole thing for a similar task

With Claude projects:

1. You set up the project once
2. You select the project in the chat (the + icon)
3. You write a short request
4. You get the result taking the whole context into account

Claude Projects vs ChatGPT: the key differences

📊 Comparison table

Feature	Claude Projects	ChatGPT Custom GPTs
File access	Direct upload + Google Drive/GitHub	Direct upload only
System instructions	Detailed, flexible	Limited
Usage	From any chat with a click	A separate interface
Number of files	Up to 100 files	~20 files
File size	Up to 100MB each	Up to 25MB each
Integrations	Google, GitHub, Notion	Limited
Context	Kept between sessions	Lost when switching

🏆 The advantages of Claude Projects

1. Ease of use:

• No need to switch to a separate interface
• Pick the project right in the chat
• Instant access to the whole knowledge base

2. Flexible setup:

• Detailed system instructions
• The ability to update files
• Combining with Claude’s other features

3. Professional integrations:

• A direct connection to Google Drive
• Sync with GitHub repositories
• Working with corporate data

Creating and setting up projects

🚀 Creating a project step by step

Step 1: Create the project

1. Open Claude’s sidebar
2. Click “Create project”
3. Enter a name (e.g. “YouTube Channel — Elton Labs”)
4. Add a project description

Step 2: System instructions

# Role: assistant for the Elton Labs YouTube channel
 
## Channel context:
- Topic: No-code Tools, AI, Obsidian
- Audience: 7000 subscribers
- Style: educational, practical
- Goal: help people automate their work without coding skills
 
## Tasks:
1. Creating video scripts
2. Analysing trends in the niche
3. Generating content ideas
4. Optimising metadata (titles, descriptions, tags)
5. Competitor analysis
 
## Answer style:
- Structured answers with headings
- Concrete examples and cases
- An emphasis on practical application
- Modern trends and tools

Step 3: Adding files

• Upload past video scripts
• Add research on the topic
• Attach spreadsheets with channel analytics
• Connect Google Drive with the content plan

📁 File types for projects

Documents:

• PDF files (books, reports, research)
• Word documents (scripts, plans)
• Google Docs (via the integration)
• Markdown files (documentation)

Data:

• CSV files (analytics, statistics)
• Excel spreadsheets (plans, calculations)
• JSON data (API responses, configs)

Media:

• Images (diagrams, screenshots)
• Presentations (slides for videos)

Code:

• Files from GitHub repositories
• Automation scripts
• Config files

Examples of ready-made projects

🎬 YouTube Channel — Content Creator

System instructions:

# YouTube Content Assistant
 
## Specialisation: No-code Tools, AI, Obsidian
 
## Main tasks:
1. **Trend analysis** — finding current topics in the niche
2. **Creating scripts** — structured video scripts
3. **SEO optimisation** — titles, descriptions, tags for YouTube
4. **Competitor analysis** — what other channels are making
5. **Content plan** — ideas a month ahead
 
## Knowledge base:
- An archive of past videos and their metrics
- Research on the target audience
- Trends in the No-code and AI industry
- Templates for different content types
 
## Working style:
- Practical advice with examples
- Structured delivery of information
- A focus on implementation, not theory
- Modern language that beginners understand

Project files:

• youtube-analytics.csv — channel metrics
• content-plan-2025.xlsx — content plan
• competitor-research.pdf — competitor analysis
• audience-survey.json — subscriber survey

🧠 Psychological Analysis

System instructions:

# Personal psychologist-analyst
 
## Role:
An experienced psychologist-consultant for self-analysis and growth
 
## Methodology:
1. **Active listening** — understanding the emotional state
2. **A cognitive-behavioural approach** — identifying thought patterns
3. **Practical techniques** — concrete exercises for change
4. **Long-term growth** — strategies for personal development
 
## Working format:
- Defining the problem or request
- Analysing the possible causes
- Practical recommendations (2-3 options)
- Homework to reinforce it
- 5 clarifying questions to understand the context
 
## Principles:
- Empathy and support
- Non-judgemental
- A focus on solutions, not problems
- Developing the client's independence

Knowledge base:

• psychology-books.pdf — professional literature
• personal-journal.md — a self-observation journal
• techniques-library.docx — a collection of psychological techniques
• case-studies.xlsx — analysis of past sessions

💻 Website Optimisation — Technical SEO

System instructions:

# SEO and web optimisation expert
 
## Specialisation:
Technical SEO, UX optimisation, website performance analysis
 
## Areas of expertise:
1. **Core Web Vitals analysis** — LCP, FID, CLS metrics
2. **Structured data** — Schema.org markup
3. **Performance** — load optimisation
4. **Mobile adaptation** — a Mobile-First approach
5. **Content strategy** — SEO-friendly texts
 
## Tools:
- Google PageSpeed Insights analysis
- Search Console recommendations
- Lighthouse audits
- Competitive SERP analysis
 
## Approach:
- Data and metrics above all
- Practical steps, not theory
- Prioritising tasks by ROI
- Measurable results

Connected resources:

• Google Analytics data
• Search Console reports
• Lighthouse site audits
• Keyword research

🎯 Principles for organising projects

1. One project = one area of work

✅ Right:
- "Elton Labs YouTube channel"
- "Personal psychotherapy"
- "Website SEO audits"

❌ Wrong:
- "Various tasks"
- "General assistant"
- "All projects at once"

2. Clear system instructions

# The structure of an effective instruction:
 
## Role and context
Who you are and what area you work in
 
## Specific tasks
3-7 main functions, no more
 
## Knowledge base
Which files to use and how
 
## Working style
How to respond, what tone, what structure
 
## Constraints
What NOT to do, the boundaries of competence

3. Regularly updating the knowledge base

• Add new files as they appear
• Remove outdated information
• Update the instructions when the tasks change

📈 Project effectiveness metrics

Measure the results:

Time saved:

• Before the project: 15 minutes explaining the context
• With the project: 2 minutes per request
• Saving: 87% of the time

Quality of results:

• Relevance of answers: 60% higher
• Style consistency: a stable 90%+
• The need for clarifications: down by 70%

Practical tracking:

# Project effectiveness log
 
Project: YouTube Channel
Period: January 2025
 
Tasks completed: 25
Average time per task: 3 minutes
Result quality (1-10): 8.7
Number of iterations: 1.2
 
Time saved: 4.5 hours a month
Project ROI: 300%

Advanced techniques

🔄 Multi-project workflows

Scenario: creating educational content

Stage 1: Research

Project: "Marketing research"
Task: find trending topics in the No-code niche

Stage 2: Creating content

Project: "YouTube channel"
Task: create a script on the found topic

Stage 3: SEO optimisation

Project: "Content SEO"
Task: optimise the metadata for YouTube

🎨 Combining with other features

Project + Artifacts:

• Create interactive materials in the project’s context
• Save the results as artifacts for reuse

Project + Styles:

• Set a default style for each project
• Use situational styles within projects

Project + Extended thinking:

• Turn on thinking mode for complex analytical projects
• See the full logic behind the decisions

🔧 Troubleshooting projects

Common problems and solutions:

Problem: Claude forgets the project context

Solution:
1. Make sure you selected the project via "+"
2. Remind it of the project at the start of the request
3. Update the system instructions

Problem: files won’t upload

Solution:
1. Check the file size (max 100MB)
2. Use supported formats
3. Try uploading via Google Drive

Problem: low-quality answers

Solution:
1. Refine the system instructions
2. Add examples of good results
3. Update the knowledge base with relevant files

Claude Projects is a professional tool for automating recurring intellectual tasks. Unlike ChatGPT Custom GPTs, Claude projects offer more.

A properly set-up project can save 10-20 hours a month and significantly improve the quality of your results. The time invested in creating a project pays off within a week of active use.

📺 Practical examples: Setting up Claude projects for automation

---

Related articles:

• Claude styles: building your own assistants
• Deep analysis from Claude: the triple mode
• Claude Artifacts: games, quests, assistants

Tags: claude projects automation productivity workflow ai