tokenwise
Lets your agent choose the cheapest, most efficient AI model for each task. Helps save money on AI processing costs.
What it does
TokenWise helps save money on AI usage by choosing the cheapest AI model that can get the job done. It logs how much each task cost and lets you test which model works best for you.
When to use it
- ✓To cut costs without losing response quality.
- ✓To check if a simpler model is good enough for your tasks.
- ✓To understand where AI money is being spent.
Example of what to ask your AI
“I want to test if the Haiku model can rewrite this paragraph to be simpler.”
Want more? Open the sections below.
How to use
The easiest way: copy the message below and paste it into Claude, ChatGPT, Cursor or any AI. It reads the skill and follows the instructions.
Skill link
Ready-to-paste message
How it works
TokenWise analyzes your task and sends it to the most suitable and cost-effective AI model. It records the exact cost of each task and allows you to compare different models to ensure you're using the best option.
Installation
Easiest way: copy the skill link and paste it into Claude, ChatGPT or Cursor asking it to follow the instructions.
In the terminal (Claude Code, Codex, Cursor), in your project folder:
npx skills add sickn33/agentic-awesome-skills --skill tokenwiseOr download just the file:
curl -L https://raw.githubusercontent.com/sickn33/agentic-awesome-skills/main/skills/tokenwise/SKILL.md -o SKILL.mdTechnical content (original SKILL.md)
Original text by the author, straight from GitHub.
TokenWise — Measurement-Driven Model Router
Overview
A Claude Code skill that auto-routes subtasks to the cheapest model that can handle them (Haiku for grunt work, Sonnet for scoped reasoning, Opus only for synthesis), then logs every routed task to a local NDJSON with real token + cost numbers. Includes an A/B test subcommand that runs the same task across multiple tiers and scores quality, so the routing decisions are verified against the user's real workload — not estimated.
Anthropic's own bug tracker (Issue #27665) reports 93.8% of Max-subscriber Claude Code tokens flow to Opus. Existing routers (claude-router, wshobson, VoltAgent) either pin models statically or route by vibes-based heuristics with no measurement. TokenWise fills the measurement gap.
When to use
- Cutting Claude Code token spend without sacrificing output quality
- Validating whether Haiku/Sonnet is "good enough" for a specific task class before trusting auto-routing
- Auditing where Opus tokens are actually being burned
- Logging per-session cost data for finance or chargeback
Subcommands
/tokenwise:install— guided installer with diff preview, automatic backups, and--dry-runmode/tokenwise:report— per-session token + cost summary vs all-Opus baseline/tokenwise:summary [--week|--month|--all]— historical aggregate with trend/tokenwise:ab "<task>"— A/B test the same task at multiple tiers, generates a markdown comparison/tokenwise:undo— restore CLAUDE.md / settings.json from backup
Routing taxonomy
| Tier | Model | Task class |
|---|---|---|
| Mechanical | Haiku 4.5 | file reads, grep, format, rename, simple edits, doc lookups |
| Scoped reasoning | Sonnet 4.6 | single-file refactor, scoped research, test writing |
| Synthesis | Opus 4.7 | architecture decisions, multi-file refactor, security review |
Safety caps:
- Haiku never spawns further subagents
- Max spawn depth = 2
- Subagents that need a smarter model return to parent — they never escalate on their own
- Tasks under 100 chars with no file context run inline (subagent overhead > savings)
- Subagent context >30k tokens bumps a tier
Privacy
Zero telemetry. All logs in .tokenwise/log.ndjson local to the project. Task descriptions truncated to 80 chars and stripped of file contents before logging. No analytics endpoint exists in the source.
Install
In any Claude Code session:
/plugin marketplace add CodeShuX/tokenwise
/plugin install tokenwise@tokenwise
Then run /tokenwise:install and follow the guided prompts.
Limitations
- Token counts approximate to ±2% vs Anthropic billing
- A/B test mode costs extra tokens (one task × N tiers) — intentional one-time validation
- Anthropic-only by design (use LiteLLM or OpenRouter for cross-vendor)
- Subagent
model:param has known silent-fail bugs on some Claude Code builds — skill probes for this at install and refuses to configure if routing is broken
Source
- Repo: https://github.com/CodeShuX/tokenwise
- License: MIT
- Author: CodeShuX
FAQ
How does TokenWise choose the right model?
It uses a predefined list: simple, quick tasks go to Haiku, tasks needing a bit more reasoning go to Sonnet, and complex tasks like planning software architecture go to Opus. It can also upgrade the model if a task turns out to be more complicated than expected.
Does TokenWise send my data anywhere?
No, it's completely private. All cost information and tasks are saved only on your computer, and sensitive data is protected.
What is the A/B test?
It's a feature that lets you run the same task on different AI models simultaneously. This helps you see which model delivers the best quality for your specific use, ensuring that savings don't harm the outcome.
Details
Original description by the author
Measurement-driven model router for Claude Code. Routes Haiku/Sonnet/Opus per task class, logs every routed task with real $ numbers, and A/B tests cheaper tiers before you trust the savings.
Popularity
- GitHub stars
- 48k
- Ranking position
- #4689
- License
- MIT
- Updated
- 10/15/2026
Comments
- …
More skills from this repository
00-andruia-consultant
Helps your agent start or improve AI projects. It analyzes code, suggests tools, and creates action plans in Spanish.
sickn33/agentic-awesome-skills
007
Lets your agent perform security audits and threat modeling. It's useful for protecting projects against attacks and vulnerabilities.
sickn33/agentic-awesome-skills
10-andruia-skill-smith
Helps your agent create or improve AI tools. It defines how tools work, writes documentation, and validates the structure.
sickn33/agentic-awesome-skills
20-andruia-niche-intelligence
Lets your agent create a summary about a specific industry. It's useful for understanding the context, rules, usage patterns, and risks of a project, using reliable sources.
sickn33/agentic-awesome-skills
Similar skills
agent-architecture-audit
Makes your agent check if everything is working correctly. It finds and fixes problems in AI systems, ensuring they perform as expected.
affaan-m/ECC
agent-eval
Helps your agent compare different AI coding tools. It measures which one is faster, cheaper, and more reliable for specific tasks.
affaan-m/ECC
agent-harness-construction
Makes your agent better understand how to perform tasks. It optimizes the tools and instructions the agent uses to increase its chances of success.
affaan-m/ECC