CNB author instance — reading feature flag config, checking content fragment status
🏢 CNB AEM Cloud21 MCP callsLast: 12m ago
Wrangler CLI
○ Idle
Last run: wrangler deploy --env production on email.experiencelab.app
Last active: 2h ago
Gmail MCP
○ Idle
Last used: drafted stakeholder update re: AEP Web SDK migration timeline
Last active: yesterday
✅
2
Active Now
Running this session
⌨️
5
Code Projects
Apps & tools being built
🤝
4
Work Projects
CNB & client work
💬
6
Chats
Personal conversations
✔️
42
Total Tasks
Across everything
🪙
74K
Est. Tokens
All projects combined
All Activity
NAME
TYPE
LAST ACTIVE
TASKS
TOOLS USED
EST. TOKENS
STATUS
Loading…
💬
My Chats
Regular back-and-forth conversations with Claude. Use the sidebar to browse all chats or filter by category.
💬
6
Total Chats
All conversations
🔥
3
Used Today
Opened in last 24hrs
✅
9
Tasks Done
Completed across chats
📝
—
Words Exchanged
Estimated this session
🏡
3
Personal
Gazebo, travel, stocks…
💼
3
Research
CA Gov, investing, news
Breakdown by Category
🏡
Personal
3 chats
Home projects, travel, and lifestyle conversations. Quick questions and planning sessions — no files or long context needed.
✓ Gazebo build planning
✓ Las Vegas trip planning
✓ Spicy chai recipes
Loading…
🔍
Research
2 chats
Deep-dive research and analysis sessions — stocks, politics, geopolitics. These tend to be longer with lots of back-and-forth.
✓ Investing & stock research
✓ 2026 CA Governor race
— Home improvement ideas
Loading…
⚡
Quick Q&A
1 chat
Short, one-off conversations for quick answers, ideas, or help with something specific. In and out fast.
✓ Short focused exchanges
✓ Brainstorms & one-off tasks
✓ Quick drafts & edits
Loading…
All Chats
All Conversations
—
CHAT NAME
CATEGORY
LAST ACTIVE
TOPICS
EST. TOKENS
Loading…
💡 Chat vs Project — what's the difference?
A Chat is a simple conversation — like texting. Ask, Claude answers. Great for quick questions or one-off tasks.
A Project is a folder for bigger ongoing work with files, instructions, and memory. Like a dedicated work binder.
⌨️
—
Coding Projects
Apps & tools being built
🤝
—
Work Projects
Job tasks & deliverables
✅
—
Total To-Dos
Tasks across all projects
📊
—
Words Used
Estimated across projects
🧩 What are "Special Abilities"?
These are pre-loaded instruction sets that teach Claude how to do specific jobs really well — like building a PowerPoint deck, writing Adobe Analytics code, or creating a Word document. Think of them like apps installed on a phone. You have 160+ of them active.
All Projects
Your Active Projects
—
Loading…
Special Abilities Unlocked 🧩
—
To-Do Lists by Project Type
● Coding To-Dos
0
● Work To-Dos
0
🌐
Websites
Every web app, tool, dashboard, and landing page built with Claude. React apps, single-file HTML tools, full SaaS platforms, and brand sites — all in one place.
✅
1
Live
Deployed & public
💾
7
Local
Built, not deployed
⚛️
4
React
Component-based
📄
4
HTML/JS
Single-file sites
All Websites
Website Registry
8 sites
NAME
STACK
BUILT
STATUS
DESCRIPTION
LINK
Loading…
📄
Files & Documents
Every deliverable, deck, doc, and packaged file created with Claude. PowerPoints, Word docs, Markdown plans, and installable skill files — all tracked here.
📊
2
PPTX
Slide decks
📝
2
DOCX
Word documents
📋
1
Markdown
Plans & specs
📦
4
.skill
Packaged skills
All Files
File Registry
6 files
NAME
TYPE
PROJECT
DESCRIPTION
NOTES
DOWNLOAD
Loading…
🧩
My Skills
Custom instruction sets you've built that teach Claude how to do specific jobs. Each skill is a SKILL.md file stored in your Claude environment — Claude reads it automatically when the task matches. Think of them as reusable SOPs for Claude.
🧩
14+
Skills Built
10 installed + project skills
⭐
4
A-Grade
Production ready
🔧
4
B-Grade
Good, needs polish
🚧
1
C-Grade
Early / narrow scope
📁
Project Skills Not Fully Visible
Skills inside your "1-Person Marketing Business" Claude Project (newsletter name picker, extended affiliate builder, and others) can't be read from outside that project. Rows marked 📁 1-Person Marketing Business are placeholders. To grade them: open that project → copy the SKILL.md content → paste it here.
All Skills
Skill Registry
10+ skills
SKILL NAME
CATEGORY
CREATED
SIZE
GRADE
DESCRIPTION
IMPROVEMENTS NEEDED
DOWNLOAD
Loading…
📊 Grading Rubric
A
Production Ready
Clear trigger phrases, strict scope boundaries, full step-by-step workflow, edge cases handled, refusal cases defined, output format specified.
B
Good, Needs Polish
Works well in most cases but missing some edge cases, triggers could be sharper, or output format not fully locked down.
C
Early Stage
Core idea is solid but scope is narrow, workflow is shallow, or instruction depth isn't enough for consistent results.
D
Needs Rework
Instructions are vague, no clear scope boundaries, or Claude produces inconsistent/unpredictable outputs when following it.
⚖️
Compare All Tools
See all three tool types side by side. MCPs plug into Claude automatically · APIs are called by your code · CLIs are terminal programs. Use the sidebar to view each type individually.
🔌
13
MCPs
Claude plug-ins
🌐
4
APIs
Web services
⌨️
3
CLIs
Terminal tools
✅
4
MCPs Active
Used this session
📞
—
MCP Calls
This session
🏆
CF
Most Used
By call volume
Side-by-Side Comparison
MCP
Plug-ins
13 connected
Claude uses these automatically during a conversation. No commands needed — just connect once and Claude decides when to use them.
✓ Works inside Claude chat
✓ No setup per conversation
✓ Claude picks when to use it
✗ Not for running code locally
✗ Requires a server to be running
Loading…
API
Web Services
4 in use
Your code calls these over the internet. Think of it as your app sending a question to a smart website and getting an answer back.
✓ Called from your code or apps
✓ Works anywhere with internet
✓ Powers ExperienceLab & APEX
✗ Costs money per use (tokens)
✗ Needs an API key to authenticate
Loading…
CLI
Terminal Tools
3 installed
Programs you type commands into on your computer. You install them once, then control them by typing in a terminal window.
✓ Runs on your local machine
✓ Fast and precise for developers
✓ Deploy, build, and manage files
✗ Must be installed first
✗ Requires typing exact commands
Loading…
🔌
MCPs — Claude Plug-ins
MCPs are apps that plug directly into Claude. Once you connect one in Settings, Claude uses it automatically during conversations — no typing commands, no extra steps. Like giving Claude a new superpower.
🔌
13
Total Connected
In your account
✅
4
Active This Session
Cloudflare, AEM, Gmail, Drive
😴
9
Idle
Connected but not used today
📞
—
Calls This Session
Tool invocations
All 13 Connected MCPs
Loading…
🌐
APIs — Web Services Your Code Calls
An API is a web address your code talks to directly. You send it a request (like "give me a response to this message") and it sends one back. APIs power all your apps — ExperienceLab, APEX, and CNB integrations all run on these.
🌐
4
APIs in Use
Active endpoints
✅
4
All Active
All online this session
🔑
2
Auth Types
Bearer token, API key
💰
Pay/use
Billing Model
Per token or per call
All APIs
Loading…
⌨️
CLIs — Terminal Programs
A CLI is a program you install on your computer and control by typing commands. Think of it like a power tool — fast and precise, but you have to set it up first and know the commands. You use these to deploy code, manage cloud services, and run Adobe workflows.
⌨️
3
CLIs Installed
On your machine
✅
2
Active
Claude Code & Wrangler
😴
1
Idle
Adobe AIO CLI
🚀
Local
Where It Runs
On your computer
All CLIs
Loading…
💡
Tips & Tricks
22 tips to help you get more out of Claude — save tokens, stay organized, work faster, and get better results. Filter by category below.
🪙
What Is a Token?
A token is the fundamental unit Claude uses to read and write text — roughly
¾ of a word, or about 4 characters on average.
Every API call is priced and metered in tokens. There are five distinct token types,
each with a different role, cost, and behavior. Understanding them helps you optimize your usage —
especially when running long sessions, building tools with system prompts, or using prompt caching.
1 token ≈ 4 chars1 token ≈ ¾ word200K token context window~150K words fits in context
The 5 Types of Tokens Explained
📥
Input Tokens
You → Claude
Everything Claude reads to generate a response. This includes your messages, the system prompt, any uploaded documents, tool results fed back in, and the entire conversation history replayed each turn.
Your typed message each turn
System prompt (operator instructions)
Full conversation history (re-sent every turn)
PDFs, docs, and images you upload
Tool/MCP results returned to Claude
Session share—
📤
Output Tokens
Claude → You
Every token Claude generates in its response — text, code, structured data, reasoning, and tool calls. Output is metered per-token and is typically the most expensive token type per million on the API.
All written response text
Generated code blocks
Structured JSON or data
Tool invocation payloads (e.g. MCP calls)
Extended thinking reasoning tokens
Session share—
💾
Cache Write Tokens
Saved for later
When Claude writes a prompt cache checkpoint, these are the tokens being saved. Cache writes cost 25% more than normal input — it's the one-time "storage cost." Writes are triggered by the cache_control parameter in the API.
When a cached prefix is reused, Claude reads it at 90% cheaper than normal input tokens. This is the big efficiency win — long system prompts and skill definitions are served from cache on every turn after the first write.
System prompt served from cache (90% off)
Skill SKILL.md files reused across turns
Memory/userMemories block each turn
All 200+ tool definitions (MCP schemas)
Your userMemories block = ~8K tokens cached
Session share—
How Your Tokens Are Broken Down
Your Token Mix (This Chat)
Loading…
💰 How to Save Tokens
💡
Cache your system prompt
At 200K ctx, caching a 50K system prompt saves ~90% on that portion every turn. In heavy sessions this cuts cost by 60–80%.
📐
Minimize history replay
Each turn replays the full conversation as input. Summarize old turns or use cache_control to cap input growth in long sessions.
🎯
Use Sonnet over Opus for iteration
Sonnet 4.6 is ~5× cheaper per token than Opus while handling 90% of tasks. Use Opus only for complex reasoning that needs it.
⚡
Cache read hits are nearly free
In this session, your system prompt (skills, memory, tools) is served from cache. That's ~30K–60K tokens per turn at 10¢/M instead of $3/M.
💵 What Each Model Costs
Price Per 1 Million Tokens
claude.ai/pricing
Model
Input
Output
Cache Write
Cache Read
Context
Relative Cost
Current Subscription
Claude Max5×
$125.00 / month · billed monthly via App Store (web: $100/mo direct)
● YOU ARE HEREMax 5× · $125/mo App Store
↑ Upgrade availableMax 20× · $250/mo App Store
💡 Save $25/mo: Subscribing directly at claude.ai costs $100/mo vs $125/mo in the App Store. Apple's ~25% platform fee is the only difference — same plan, same features, same usage.
⚡
Usage vs Pro
5× more
💬
Est. messages/5hr
~150–200
🔌
MCP Apps
Unlimited
🧠
Memory
Persistent
🖥️
Claude Code
Full access
🔝
Priority Access
Peak demand
5-Hour Window Usage
How Much of This 5-Hour Block You've Used
ESTIMATED
Loading…
Messages This Window
—
of ~200
Tokens Consumed
—
est. cap
Window Resets In
—
5 hr window
Equivalent API Cost
—
vs $125/mo
⚠️ Estimated only — Anthropic doesn't expose live usage data via API. These figures are calculated from tokens tracked in this dashboard session plus a baseline estimate for your typical daily usage pattern. To see exact limits, visit claude.ai → Settings → Billing.
💰 Are You Getting Your Money's Worth?
Here's what this would cost if you paid per message instead of a flat monthly fee:
Estimated Monthly Cost Without Your Plan
—
You pay $125/mo · saving —
📅 Your Billing Info
Plan
Billed monthly, App Store
Max 5×
Monthly charge$125.00
Web price (direct)$100.00
Apple platform fee$25.00
Daily cost~$4.17
Cost per window (5hr)~$0.87
Cost per message est.~$0.004
💡 Save $300/yr
Switch to claude.ai direct billing at $100/mo. Same plan, Apple just keeps $25.
What You Get Each 5-Hour Window
⚡
5×
Usage vs Pro
Resets on rolling 5-hour window
💬
~200
Messages / 5 hrs
Varies by message length & model
🧠
200K
Context Window
All models, per conversation
🖥️
Full
Claude Code Access
Higher rate limits than Pro
🤖 Which Claude Models You Can Use
Available Models
5 models
✅ Everything Included in Your Plan
📈 If You Ever Need to Upgrade
Pro vs API — What's the Difference?
claude.ai Max 5× ($100/mo web · $125/mo App Store)
YOUR CURRENT PLAN
Flat monthly fee — unlimited token consumption within the 5-hour rolling usage window. No per-token billing.
You get ~150–200 messages per 5-hour window, priority model access, full Claude Code, all MCPs, memory, and skills.
Given the volume of complex sessions you run (ExperienceLab, CNB, coding), Max 5× is well justified — the equivalent API spend
would be hundreds to thousands per month at token rates.
Worth upgrading to Max 20× ($200)?
Only if you're regularly hitting the 5-hour limit mid-session. Max 20× gives 4× more runway. Claude Code and claude.ai share the same quota pool — heavy CLI use counts against your chat limit.
Anthropic API (pay-per-token)
FOR APPS YOU BUILD
Direct access at api.anthropic.com — billed per token consumed, no usage cap.
Used for ExperienceLab, APEX, AdForge, and any tool you build with claude-sonnet-4-20250514.
Completely separate billing from your Max subscription.
This dashboard's live data fetch is an API call — billed to your API account, not your Max plan.
Sonnet at $3/$15 per MTok
At API rates, a heavy ExperienceLab dev session (500K tokens) costs ~$6–10. Caching your system prompt drops that 60–80%. Still, Max 5× pays for itself fast if you use Claude daily.