skills.rest
langwatch
langwatchOFFICIAL
Explore all skills by this author →
GitHub

provider-cost-comparison

Reprice real LangWatch token usage against other model providers including cache costs.

3.5k|362|Updated Sep 9, 2023
Search Tags:#trace export#LLM cost#token pricing#provider comparison#Cost Analysis#langwatch#cache pricing
One-click install
npx skills add https://github.com/langwatch/langwatch --skill provider-cost-comparison
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: provider-cost-comparison
Source: https://github.com/langwatch/langwatch/tree/main/skills/_compiled/native/provider-cost-comparison
Command: npx skills add https://github.com/langwatch/langwatch --skill provider-cost-comparison

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires langwatch.

What problem does it solve?

Sticker-price comparisons between LLM providers ignore cache-read and cache-write pricing, so a provider with cheaper per-token rates can actually cost more on cache-heavy workloads. This Skill prices your actual month of LangWatch usage, with its real input/output/cache mix, under each candidate provider.

Core Features & Use Cases

  • Real Usage Export: Exports your actual token mix per model via the LangWatch CLI, including the cache read/write split from session events, scoped by window and origin.
  • Current Price Fetching: Fetches live provider price pages at analysis time and records cache pricing shape (write premium, storage-by-time, or no cache).
  • Repricing with Sensitivity: Reprices the same usage under each candidate with direct, no-cache-degradation, and cache-hit sensitivity rows, then writes a self-contained HTML report.
  • Use Case: Ask whether switching your coding-agent workload from one provider to another would save money, and get a report showing your $X actual spend versus $Y repriced, with the cache assumptions stated.

Quick Start

Ask the assistant to check whether another model provider would be cheaper for your LangWatch usage over the last 30 days.

Frequently Asked Questions about provider-cost-comparison

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I compare LLM provider costs for my actual usage?▼

Export your real token mix per model with the langwatch CLI, including cache read and write splits, then reprice that same usage under each candidate provider's current price card. The analysis includes direct repricing, no-cache degradation, and sensitivity across cache-hit shares.

Why do cache prices matter when comparing model providers?▼

On coding-agent workloads, cache reads are often the large majority of all input tokens, so a provider with cheaper per-token prices but no cache discounts can cost more overall. Cache pricing also comes in three shapes (write premium, storage by time, or none) that reprice differently.

How do I install and authenticate the langwatch CLI?▼

Install it once with npm install -g langwatch and run the langwatch binary directly. Personal coding-agent usage needs langwatch login --device, while team or application projects need --project <slug> on read commands.

Does the cost comparison work for both coding agents and application traffic?▼

Yes, but the two scopes must not be mixed. Scope the trace export with --origin coding_agent for coding-agent questions, or drop the flag for whole-project analysis, since analytics queries cover the entire project and have no origin filter.

What happens if the trace export hits the limit?▼

The --limit flag caps the whole export, so a window with more matches produces a partial file. The command reports both counts when truncating; raise the limit until they agree or state in the report that numbers come from a sample of N of M traces.

Can a repriced number predict my bill after switching providers?▼

No, repricing prices the same usage at different rates, and a different model changes the usage itself. The report states this once clearly and names non-price caveats like context window and capability differences without judging quality.

Related Skills

Explore More Data & Analytics Skills→
ContextJet-aiContextJet-ai
29

instrument-llm-observability

Add OpenTelemetry tracing to LLM applications capturing prompts, tool calls, token usage, latency, and cost.

Official
Intermediate
jndeweyjndewey
2

construction-loan-negotiation

Analyze construction loan agreement provisions against lender, borrower, and market benchmarks.

Community
Advanced
latestaiagentslatestaiagents
5

agent-cost-budgeting

Track token usage, enforce budget limits, and generate cost reports for AI agents.

Official
Advanced
tankygranny05tankygranny05
2

token-cost-computation

Calculate token usage totals and estimated costs from Codex CLI and Claude Code SSE JSONL logs.

Community
Intermediate