Individual skill
cost-local
Cost per million tokens on hardware you own (Ollama, llama.cpp, vLLM, LM Studio) from watts, electricity price, hardware price and measured tokens/second, and the utilisation at which local beats a hosted model. Use for local-vs-API break-even questions.
plugins/ruflo-cost-tracker/skills/cost-local/SKILL.md
Source commit 722232566ad1fc4e73ba4fefc46d77d380ce61ba
What was checked
- Matches the observed repository HEAD
- Yes
- Scanner rules are current
- No
- Repository scan complete
- No
- Skill scan complete
- No
- Observed behavior
- Not checked yet
- Last checked
- Nov 6, 2026
These are observations at the recorded commit, not a guarantee of safety. Review the source before running it.
Install this skill
Choose an agent in your Stack or copy a command below. Commands fetch the pinned commit and copy only this skill directory. Existing destinations are refused. Copying a command does not install anything.
An install command is unavailable: current rules, complete repository and skill scans, an eligible behavior level and a permitted license are required.
Repository context
- Repository stars (not skill ratings)
- 81,243
- Repository last updated
- Nov 7, 2026
- License
- MIT
Observed agent integration: No specific integration evidence. Standard SKILL.md installation does not imply verified compatibility.
- swarm Repository topic
- agentic-ai Repository topic
- agentic-framework Repository topic
- agentic-workflow Repository topic
- autonomous-agents Repository topic
- codex Repository topic
- mcp-server Repository topic
- multi-agent Repository topic
- ai-assistant Repository topic
- multi-agent-systems Repository topic
- engineering Derived category
Files and excerpt
File metadata and permitted, redacted excerpts come from the recorded scan. Source links are pinned to that commit.
Read excerpt
# Cost Local
```bash
node ${CLAUDE_PLUGIN_ROOT}/scripts/local-cost.mjs --tok-per-s 80 --busy 0.25 --compare claude-haiku-4-5
```
Inputs (all the user's): `--watts-idle`, `--watts-active`, `--kwh`, `--hw-price`, `--life-years`, `--resale`. `--tok-per-s` is required and must be a **measured** generation speed.
## Rules
- Cost = **fixed** (amortisation + idle power, paid busy or not) + **marginal** (extra watts while generating). Utilisation decides the answer: always show the table at 5/25/50/100% busy.
- It is an estimate from the inputs; say so. Nothing is measured.
- Compare against a same-capability hosted model, not a frontier one; price alone is a poor reason to buy a GPU for sporadic use.
- Read real counts from the server (Ollama `eval_count`/`prompt_eval_count` on the native API, llama.cpp `timings`, vLLM `/metrics`), not the lossy OpenAI-compatible `usage`, and count model reloads (`load_duration`) as overhead.