Releases·Tool Release
Claude Code adds token-visibility updates to /cost, /usage, and /tasks commands
An X post announces new token-visibility features for Claude Code: /cost gains a prompt-cache line, /usage breaks down /loop tasks by runs and tokens, and /tasks shows which model and effort level each subagent ran on. The changes respond to user requests for clearer insight into where tokens go, helping developers spot costly loops and audit subagent configurations.
495 engagement1 sourceFri, Aug 28, 2026, 11:27 PM
AI Overview
- /cost now includes a prompt-cache line item
- /usage breaks down /loop tasks by runs and tokens to reveal which loops burn many tokens
- /tasks shows which model and effort level each subagent ran on
- Changes were made in response to user requests for more visibility into token spend
Entities
Related
Social·Community
Claude Code users investigate early usage-limit hits: distinguishing context-window fill from subscription exhaustion
A Reddit discussion asks practitioners who investigated Claude Code consuming usage allowances faster to share what evidence connected the increased usage to a fixable cause. The thread distinguishes context-window pressure from subscription usage exhaustion and proposes a diagnosis workflow using /context and /usage.
A Reddit discussion asks practitioners who investigated Claude Code consuming usage allowances faster to share what evidence connected the increased usage to a fixable cause. The thread distinguishes context-window pressure from subscription usage exhaustion and proposes a diagnosis workflow using /context and /usage.
4 engagement1 sourceredditMon, Sep 28, 2026, 02:07 PM
Social·Community
Developer breaks down token spend in AI agent sessions
A developer analyzed token usage in long-running AI agent sessions, finding that most tokens go to resent context, tool results, and reasoning, with only a small fraction used for final output. The analysis suggests much of this overhead does not require a frontier model, prompting discussion on tracking agent token costs.
A developer analyzed token usage in long-running AI agent sessions, finding that most tokens go to resent context, tool results, and reasoning, with only a small fraction used for final output. The analysis suggests much of this overhead does not require a frontier model, prompting discussion on tracking agent token costs.
22043 engagement4 sourcesredditMon, Aug 17, 2026, 03:23 PM
Products·launched
Token quota and reset-time tracker for LLM subscription plans
A reference page tracking free and paid token quotas and reset schedules for OpenAI, Anthropic, and Google models, plus an LLM product-news log. It helps developers running agents on subscription plans understand their real cost ceiling from reset timing, not sticker price. Data is maintained as public files in git.
A reference page tracking free and paid token quotas and reset schedules for OpenAI, Anthropic, and Google models, plus an LLM product-news log. It helps developers running agents on subscription plans understand their real cost ceiling from reset timing, not sticker price. Data is maintained as public files in git.
2 engagement1 sourceredditFri, Sep 18, 2026, 12:46 PM
Social·Community
Developers discuss how to measure cost per agent run in production
A developer asks how teams calculate the cost of a single agent run in production, noting that runs span multiple model calls, tools, containers, retries, and hours of execution. The question highlights the difficulty of moving beyond simple token-based cost tracking to full infrastructure allocation.
A developer asks how teams calculate the cost of a single agent run in production, noting that runs span multiple model calls, tools, containers, retries, and hours of execution. The question highlights the difficulty of moving beyond simple token-based cost tracking to full infrastructure allocation.
37 engagement2 sourcesredditMon, Aug 10, 2026, 12:03 PM
Social·Community
Developers ask how to track AI token spend by model, tool, or repo
A developer on Reddit asks how to track AI coding agent token usage when company caps are in place, noting that Copilot shows its own numbers but not which MCP servers or repos drive costs. The post seeks community advice on breakdowns by model, tool, repo, or teammate comparison.
A developer on Reddit asks how to track AI coding agent token usage when company caps are in place, noting that Copilot shows its own numbers but not which MCP servers or repos drive costs. The post seeks community advice on breakdowns by model, tool, repo, or teammate comparison.
20 engagement1 sourceredditTue, Jul 28, 2026, 11:29 PM