Feature Request: Restore Time & Token Indicators + Add Context Usage Display
Status Fixed / completed
Maintainer reply None cached
Workaround ✓ Mentioned in thread ↓
Activity 14 comments · opened Sep 4, 2025 · closed Aug 17, 2026
Description
Claude Code needs two types of improvements for better user experience:
Restore Previously Removed Features:
- Time elapsed indicator - Shows how long Claude Code has been working on the current task
- Token consumption counter - Displays tokens used for the current operation
Add New Feature:
- Context usage percentage - Shows current context window utilization
Why These Matter
- Time tracking helps users understand task complexity and duration patterns
- Token visibility enables users to optimize prompts and understand cost implications for different task types
- Context percentage allows proactive context management instead of reactive auto-compacting
Current Problem
Without the context usage indicator, users can't make informed decisions about when to clear or compact. This leads to:
- Continuing with tasks when context is nearly full, causing mid-task interruptions
- Unnecessary clearing when context usage is still low
- Poor user experience due to unexpected auto-compact triggers
Request
- Restore the time and token indicators that were previously available
- Add context usage percentage display
Codex currently has all of these features and they significantly improve the user experience by providing transparency and enabling better resource management.
These indicators transform context management from guesswork into informed decision-making.
Showing cached comments. Read the full discussion on GitHub ↗
13 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
The "context usage percentage display" would be great ! In Cline I like to see how much context is filled and then decide if it is a good moment to trigger compaction.
Yes, Codex, Cline and Cursor they all have this functionality, and actually Claude Code will dynamically show how much context left when it want to auto-compact, so it should be easy for them to just show the indicator all the time instead of only when you approach the limit, not sure why they don't do that and what makes them decide that.
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.
Requiring a command that consumes context in order to view context, is something I would only expect from someone who vibe coded their UX design.
I'd also like this feature. I am constantly doing
/contextmaybe 3-4 times every session which ironically eats up more context.The popularity of this repo should indicate just how in-demand this feature is.
+1
Nothing new on this?
Why is this not pushed further?
Seems like a simple feature, currently achievable externally though it would look like a great addition to the base product that could help all users plan and refine their usage accordingly 🙏🏼
Until this gets built in, Cozempic gives you token estimation and context visibility right now. The
cozempic doctorcommand shows session size, token count, and bloat sources, and the guard daemon continuously reports context utilization so you can manage proactively instead of waiting for auto-compact.pip install cozempic— auto-wires on first run. https://github.com/Ruya-AI/cozempic — happy to hear if the visibility helps with planning.Until this gets built in, Cozempic's
estimate_tokenscommand gives you a breakdown of where your context tokens are going — tool results, thinking blocks, images, metadata, etc. It helps you understand what's eating your context window so you can act proactively. Also available as an MCP tool in the plugin.pip install cozempic && cozempic estimateto try it. Repo: https://github.com/Ruya-AI/cozempic — let me know if the breakdown is useful.Anthropic has a strong incentive to make usage feel lightweight in the UI, but the default display is incomplete in offering only limited context indicators that don't fully break down input/output/tool costs or provide persistent live totals. This can lead users to underestimate spend and more importantly for model and output improvement: underthink their inputs, especially given how quickly agentic tool use and context re-reading accumulate.
Reporting or emphasizing only parts of the picture feels like omission when billing covers everything.
This seems at odds with Claude's principles of being maximally helpful and honest with users. A live token usage display with clear input/output/cache/estimated cost breakdown (similar to what many users end up hacking together, or what competitors like Cursor offer) would build far more trust and let people manage their sessions proactively.
Even small transparency wins matter for a billion-dollar company. Doing the right thing here signals that customers are held in high regard, not as an afterthought.
Tools like
/context,/usageand/costare provided, but I would like a billion dollar company to lead by example in defeating the tyranny of the default.Better examples elsewhere: Cursor gets a lot of praise for superior context usage visibility and management in comparisons (e.g., clearer indicators, better control via @mentions/rules). OpenAI's playground and dashboards often feel more upfront with counters. Third-party ecosystems (ccusage, custom statuslines) fill the gap effectively. All of these validate this issue.
In case it helps in the meantime — cozempic (github.com/Ruya-AI/cozempic) surfaces exact token usage (input + cache + output, read from the usage frames so it matches the real context) plus a context % bar and cache-hit rate via
cozempic current --diagnose. It's primarily a context-pruner, but the measurement is accurate. Free/open-source. Would genuinely appreciate feedback on whether that readout covers what you're missing here.