Usage Metrics Visibility for Max Subscribers

If you’d like to monitor usage while in Max mode (before Anthropic officially supports it), you can manually patch the CLI.

Go to your node_modules, find the claude-node directory, and open cli.js. Search for the string "With your Claude Max subscription". You’ll find a line like:

if (o3()) return "With your Claude Max subscription..."

Change it to something like:

if (o3()) return mg1()

This will allow the /cost command to return your current usage stats even while using Claude Max.

That said, there’s a minor issue: input tokens seem to increase by ~0.3k every time you run /cost, even when no other interactions happen. For example:

> /cost  
  ⎿  Total cost:            $0.0052
     Token usage by model:
         claude-3-5-haiku:  5.6k input, 181 output

> /cost  
  ⎿  Total cost:            $0.0055
     Token usage by model:
         claude-3-5-haiku:  5.9k input, 203 output

So while it’s helpful as a temporary workaround, be aware that input token count might not be entirely accurate.

That said, there’s a minor issue: input tokens seem to increase by ~0.3k every time you run /cost, even when no other interactions happen. For example: `` > /cost ⎿ Total cost: $0.0052 Token usage by model: claude-3-5-haiku: 5.6k input, 181 output > /cost ⎿ Total cost: $0.0055 Token usage by model: claude-3-5-haiku: 5.9k input, 203 output ``

Seems that it's an intended behavior since some requests are performed under the hood, but I did not verify that.

See: https://docs.anthropic.com/en/docs/claude-code/costs

Command processing: Some commands like /cost may generate requests to check status

Today, after updating Claude Code, I found an interesting clue while trying to modify the "/cost" code. It might be that Claude Code will be opened up to Pro users in the next few days.

!Image

+1, I just switched to a max plan from AWS Bedrock and miss being able to check usage metrics :(

YES PLEASE! especially since claude code auto switches off OPUS and I have no idea what the usage looked like, and if its worth switching over to my api key to use it as needed. It would be great to see what the usage is to better understand consumption for things. and I also kind of feel like running through my claude max 20x usage in a couple days is something I should either be proud of or mad about, but I cant make an informed decision about that without usage info 😅

+1 we are in a token/context optimization game here. the subagents reporting their token usage is great but now we have no easy way to see how many tokens are being used by the primary one.

+1 , we need it

Hi there everyone! (@mindhunter77 @toby-andyintel @jriskin @mattgodbolt @SigismondPhilippe @ramarnat @hkarasek @C12345678901111 @0ttik @mattbnz @BnnQ @samuel-2018 @TimHolahan @cendenta @jonsampson @cassidymichael @fitz-vivodyne @minamijoyo @dasha-512 @davidlwg @JrPribs @pendrya @JEETDESAI25 @MingyuJeon @marco-altran @nicoburniske @0xdeafbeef @andrewesweet @alex-orlovskyi @Spectrum-Ian @brittlewis12 @vkryukov @faisalburhanudin @ArturoTorresMartinez @cyreb7 @garrettg123 @vanakema @FindingJohnny @jmapfm @dmitry-semenov @cyfranT @Anarios @jgrayla)

I added @XiaTiaoQAQ's excellent patch to tweakcc, an automated patcher for Claude Code that fixes several bugs and adds a variety of features to Claude Code. If you'd like to use /cost to get your session stats with your Pro/Max subscription right now, run npx tweakcc --apply. (Note that when CC is updated you'll need to apply the patches again.)

Before:

$ npm install -g @anthropic-ai/claude-code

changed 3 packages in 1s

2 packages are looking for funding
  run `npm fund` for details

$ claude
╭───────────────────────────────────────────────────╮
│ ✻ Welcome to Claude Code!                         │
│                                                   │
│   /help for help, /status for your current setup  │
│                                                   │
│   cwd: /home/user                                 │
╰───────────────────────────────────────────────────╯

> /cost 
  ⎿  With your Claude Max subscription, no need to monitor cost — your subscription includes Claude Code usage

After:

$ npx tweakcc --apply
Applying saved customizations to Claude Code...
Configuration saved at: /home/user/.tweakcc/config.json
Found Claude Code at: /home/user/n/lib/node_modules/@anthropic-ai/claude-code/cli.js
Version: 1.0.111
Applying customizations...
Customizations applied successfully!

$ claude
╭───────────────────────────────────────────────────╮
│ ✻ Welcome to Claude Code!                         │
│                                                   │
│   /help for help, /status for your current setup  │
│                                                   │
│   cwd: /home/user                                 │
╰───────────────────────────────────────────────────╯

> /cost 
  ⎿  Total cost:            $0.0000
     Total duration (API):  0s
     Total duration (wall): 3s
     Total code changes:    0 lines added, 0 lines removed
     Usage:                 0 input, 0 output, 0 cache read, 0 cache write

$ 

This issue has been inactive for 30 days. If the issue is still occurring, please comment to let us know. Otherwise, this issue will be automatically closed in 30 days for housekeeping purposes.

I would like to reopen this one, please. The initial arguments are still valid, and with the current, larger model, it seems that we reach the limit faster than before.

Ok i think this is quite a important issue for user to keep in mind, it would be great claude team actually implement it.

I would like this as well!

You can use /stats from command line only (no ide vscode extension)

<img width="540" height="513" alt="Image" src="https://github.com/user-attachments/assets/9b197a9b-35b3-471b-9473-bc92c5690bef" />

<img width="528" height="432" alt="Image" src="https://github.com/user-attachments/assets/04d61b4f-bc39-4da6-b096-41ad3e17971c" />

+1 for this feature!

My use case: I use Claude Code on two machines - home and office. Currently there's no way to see combined usage across both, and I can't tell which machine consumed more of my weekly limit.

As a Max subscriber, I'd love to see:

• Daily granular breakdown of token usage
• Per-device breakdown (which machine used how much)
• Historical data that doesn't get pruned

Tools like ccusage can parse local JSONL files, but data is siloed per machine. Official support for detailed usage metrics would be incredibly helpful.

+1 this would be awesome

+1 My use case is similar to this one.

I would like to be able to at least run claude --usage via cli. I could wrap it in a percentage and display it in my OS status bar. Maybe even claude --usage --format=json.

+1 for this feature! My use case: I use Claude Code on two machines - home and office. Currently there's no way to see combined usage across both, and I can't tell which machine consumed more of my weekly limit. As a Max subscriber, I'd love to see: Daily granular breakdown of token usage Per-device breakdown (which machine used how much) * Historical data that doesn't get pruned Tools like ccusage can parse local JSONL files, but data is siloed per machine. Official support for detailed usage metrics would be incredibly helpful.

👋🏻
@mindhunter77 @toby-andyintel @jriskin @mattgodbolt @SigismondPhilippe @ramarnat @hkarasek @C12345678901111 @0ttik @mattbnz @BnnQ @samuel-2018 @TimHolahan @cendenta @jonsampson @cassidymichael @fitz-vivodyne @minamijoyo @dasha-512 @davidlwg @JrPribs @pendrya @JEETDESAI25 @MingyuJeon @marco-altran @nicoburniske @0xdeafbeef @andrewesweet @alex-orlovskyi @Spectrum-Ian @brittlewis12 @vkryukov @faisalburhanudin @ArturoTorresMartinez @cyreb7 @garrettg123 @vanakema @FindingJohnny @jmapfm @dmitry-semenov @cyfranT @Anarios @jgrayla) @zachlagden @uju777 @hugoblanc @robertphyatt

While we wait for the official feature, for those who want usage metrics now without patching Claude Code, I built a SessionEnd hook that auto-displays session analytics on exit.

What it shows (15 configurable sections):

• Model usage per model (requests, input/output tokens)
• Cache hit rate (read / created)
• Estimated cost (via ccusage or built-in pricing table)
• Tool calls breakdown with success/error rates and error details
• Files touched (read/edited/created)
• Git diff summary, lines of code written
• Features used (MCP servers, agents, skills, teams)
• Duration (wall/active), turns, exit reason, conversation ratio

!session-summary-v3

Uses the official hooks API (no binary patching, survives CC updates). Complementary to tweakcc's /cost patch: different approach (auto-display vs on-demand), deeper metrics (15 sections vs 5 lines), no modification of Claude Code.

All sections toggleable via env vars or a CLI config tool. Includes JSONL logging for cross-session analysis.

Source + docs: session-summary.sh | CLI config tool | Full documentation

Part of the Claude Code Ultimate Guide (120+ templates, hooks, and examples).

Update: session-summary is now available as a Claude Code plugin — 2 commands, zero manual config:

claude plugin marketplace add FlorianBruniaux/claude-code-plugins
claude plugin install session-summary@florian-claude-tools

Hooks auto-wire on SessionStart + SessionEnd, show up as read-only in /hooks. No file copying, no settings.json editing.

Marketplace repo: FlorianBruniaux/claude-code-plugins

@FlorianBruniaux awesome, thanks dude

Still waiting on native metrics? I got tired of guessing and built a CLI that parses the local JSONL logs for per-project, per-branch cost breakdowns. Set a dollar cap, it SIGTERMs the session when you blow past it. No API key, no dashboard, no trust required.

https://github.com/RoninForge/budgetclaw

Any ideas how to track usage in an org for members using standard/premium seats within a team plan?