惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

MongoDB | Blog
MongoDB | Blog
大猫的无限游戏
大猫的无限游戏
H
Help Net Security
V
Visual Studio Blog
G
Google Developers Blog
Blog — PlanetScale
Blog — PlanetScale
Last Week in AI
Last Week in AI
C
Check Point Blog
D
Docker
M
MIT News - Artificial intelligence
P
Proofpoint News Feed
博客园 - 叶小钗
博客园 - 聂微东
The Cloudflare Blog
云风的 BLOG
云风的 BLOG
Martin Fowler
Martin Fowler
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
酷 壳 – CoolShell
酷 壳 – CoolShell
MyScale Blog
MyScale Blog
WordPress大学
WordPress大学
Engineering at Meta
Engineering at Meta
腾讯CDC
S
SegmentFault 最新的问题
博客园 - 【当耐特】

Hacker News: Show HN

PurrrrrFocus: Pomodoro Timer App - App Store Workflow Engine — Multi-Step Orchestration for Bun RapidPhoto: Pro Photo Editor App - App Store GitHub - DheerG/swarms: Achieve extraordinary results with claude code across a variety of tasks SPICE simulation → oscilloscope → verification with Claude Code — Lucas Gerads Show HN: VCoding – A 5 MB native Windows IDE with no dynamic dependencies Show HN: LLMs don't hallucinate because they're bad at math, it's the format GitHub - Agent-FM/agentfm-core: AgentFM is a peer-to-peer network that turns everyday computers into a decentralized AI supercomputer. AgentFM lets you run massive AI workloads directly across a global mesh of idle CPUs and GPUs. Show HN: Tracking Top US Science Olympiad Alumni over Last 25 Years GitHub - Potarix/agent-hub: One place to talk to all your agents Show HN: Runtime security for AI agents(injection,tool abuse, data exfiltration) GitHub - dubeyKartikay/lazyspotify: Terminal Spotify client for macOS and Linux GitHub - the-banana-tool/king-louie: Easy to use GUI Personal AI Assistant. Win/Linux/Mac. Show HN I made my vacation rental bookable by AI agents–no Airbnb, 0% commission GitHub - basteez/jsf-autoreload: maven plugin to enable hot reload on jsf projects uvm32/hosts/host-gdbstub at main · ringtailsoftware/uvm32 GitHub - labsai/EDDI: Config-driven engine that turns JSON into production-grade AI agents. Multi-agent orchestration, 12+ LLM providers, MCP/A2A protocols, RAG, persistent memory, and enterprise compliance (EU AI Act, GDPR, HIPAA). Built on Quarkus. GitHub - glitchnsec/fortyone-oss: AI Executive Assistant Platform Quickstart | Alien GitHub - muxshed/shed: One stream in, or many. Every destination, simultaneously. No cloud middleman, no per-channel fees, no limits. GitHub - ocrbase-hq/ocrbase: 📄 PDF/IMG ->.MD/JSON Document OCR API for PaddleOCR and GLMOCR. Self-hostable. GitHub - impactjo/home-memory: MCP server that lets your AI assistant remember everything about your home. GitHub - Sets88/dbcls: DbCls is a powerful terminal database client that supports various databases GitHub - neptun2000/heor-agent-mcp GitHub - SeanFDZ/macmind: Single-layer transformer in HyperTalk for the classic Macintosh RollQuation: Math Puzzles - Apps on Google Play GitHub - dropbox/witchcraft Show HN: Agent-cache – Multi-tier LLM/tool/session caching for Valkey and Redis GitHub - opentalon/opentalon: OpenTalon is an open-source platform built from the ground up in Go as a robust alternative to OpenClaw LinkedIn™ 职位抓取工具 - Chrome 应用商店
GitHub - samjoch/tok: Claude Tokens Counter
samjoch · 2026-06-02 · via Hacker News: Show HN

Exact Claude token counter for the terminal. Reuses your local Claude Code login from the macOS Keychain — no ANTHROPIC_API_KEY required — and calls Anthropic's official count_tokens endpoint, so the number matches what Claude actually bills.

Build

cargo build --release
# binary at target/release/tok  (optionally: cp target/release/tok ~/.local/bin/)

Requires that Claude Code is installed and you've logged in at least once (claude). The OAuth token is read from the Keychain entry Claude Code-credentials.

Interactive mode (default)

Run with no input to open a Claude-Code-style field with a live count:

  • Type or paste your prompt — the token count and estimated input cost update as you edit (debounced).
  • Ctrl+V — smart paste: attaches an image if the clipboard holds one, otherwise inserts text. (Plain ⌘V text paste also works.)
  • Pasted images appear inline as [Image 1], [Image 2], … just like Claude Code. The placeholder is display-only — the image itself is counted as an image block, the [Image N] text is not.
  • Ctrl+X — drop the last attached image · Ctrl+L — clear everything
  • Esc / Ctrl+C — quit

Pipe / scripting mode

Any text, file, or image argument switches to a single count-and-print:

tok "Hello, world!"              # -> 16 tokens · ≈ $0.000240 input
echo "from a pipe" | tok          # reads stdin
tok -q "just the number"          # -> 11   (integer only, for scripts)
tok --file prompt.txt             # count a file's contents
tok "describe this" -i shot.png   # text + image (repeat -i for more)
tok -m claude-haiku-4-5 "hi"      # price reflects the chosen model

The cost shown is the input cost (what it takes to send this prompt once) at the model's per-token input price — count_tokens doesn't generate output, so there's no output cost to include. Prices are approximate, matched by model family (opus/sonnet/haiku).

Options

flag meaning
-f, --file <PATH> read prompt text from a file
-i, --image <PATH> attach an image (png/jpg/gif/webp), repeatable
-m, --model <ID> model to count against (counts are identical across Claude models)
-q, --quiet print only the integer
--interactive force the field even when input is piped

How it works

tok reads the claudeAiOauth.accessToken (scope user:inference) from the Keychain and POSTs your text/images to https://api.anthropic.com/v1/messages/count_tokens with the anthropic-beta: oauth-2025-04-20 header. It only sends content for counting; no completion is generated and nothing is billed for tokens.

If you see a 401, your login expired — run claude once to refresh it.