Made by@agentcostsLlm usage normalizer
@agentcosts/llm-usage-normalizer
Convert Anthropic and OpenAI usage exports, or your own app's token logs, into one usage schema with customers, tasks, cached and uncached tokens, and cost from a price table you edit. Fixes the classic mistake of counting cached tokens twice, and lists unmapped keys and unpriced models.
- Activity
- New: be the first to try it
- Installs
- 0 installs
- Likes
- 0 likes
- Clones
- 0 clones
- 1{
- 2 "name": "llm-usage-normalizer",
- 3 "owner": "agentcosts",
- 4 "version": "1.0.0",
- 5 "type": "skill",
- 6 "description": "Convert Anthropic and OpenAI usage exports, or your own app's token logs, into one usage schema with customers, tasks, cached and uncached tokens, and cost from a price table you edit. Fixes the classic mistake of counting cached tokens twice, and lists unmapped keys and unpriced models.",
- 7 "category": "data",
- 8 "license": "MIT",
- 9 "compat": ["claude-code"],
- 10 "permissions": {
- 11 "shell": ["node scripts/normalize.mjs"],
- 12 "files": {
- 13 "read": ["./data/**"],
- 14 "write": ["./data/llm-prices.csv", "./data/customer-map.csv", "./out/usage-normalized*"]
- 15 }
- 16 },
- 17 "tests": {
- 18 "cases": [
- 19 {
- 20 "name": "normalizes both example exports and maps keys to customers",
- 21 "inputs": {},
- 22 "checks": [
- 23 {
- 24 "type": "command_exits_zero",
- 25 "command": "node scripts/normalize.mjs examples/anthropic-usage.csv examples/openai-usage.csv --prices examples/prices.csv --customer-map examples/customer-map.csv --out out/usage-normalized"
- 26 },
- 27 { "type": "file_contains", "path": "out/usage-normalized.csv", "text": "timestamp,customer,task,run_id,provider,model,input_tokens,output_tokens,cache_read_tokens,cache_write_tokens,cost_usd,outcome" },
- 28 { "type": "file_contains", "path": "out/usage-normalized.csv", "text": "bluepeak,code_review,,openai,gpt-5-2025-08-07,432000,98000,180000,0,1.5425," },
- 29 { "type": "file_contains", "path": "out/usage-normalized-report.md", "text": "`key-staging` (1 row)" },
- 30 {
- 31 "type": "output_matches_schema",
- 32 "path": "out/usage-normalized-report.json",
- 33 "schema": {
- 34 "type": "object",
- 35 "required": ["rows", "total_cost", "by_customer", "unmapped_keys", "unpriced_models"],
- 36 "properties": {
- 37 "rows": { "const": 12 },
- 38 "total_cost": { "const": 11.59 },
- 39 "by_customer": { "type": "object", "properties": { "cobalt-labs": { "const": 2 }, "internal": { "const": 1.4 } } },
- 40 "unmapped_keys": { "const": { "key-staging": 1 } },
- 41 "unpriced_models": { "const": {} }
- 42 }
- 43 }
- 44 }
- 45 ]
- 46 },
- 47 {
- 48 "name": "reads app logs with OpenAI-style usage fields and flags unknown models",
- 49 "inputs": {
- 50 "data/app-log.csv": "created_at,customer_id,workflow,trace_id,model,prompt_tokens,prompt_tokens_details.cached_tokens,completion_tokens,status\n2026-10-05T12:00:00Z,acme,triage,t1,gpt-5-mini,1000000,400000,100000,success\n2026-10-05T12:01:00Z,acme,triage,t2,new-model-x,1000,0,10,failure\n"
- 51 },
- 52 "checks": [
- 53 { "type": "command_exits_zero", "command": "node scripts/normalize.mjs data/app-log.csv --prices examples/prices.csv --out out/usage-normalized" },
- 54 { "type": "file_contains", "path": "out/usage-normalized.csv", "text": "acme,triage,t1,openai,gpt-5-mini,600000,100000,400000,0,0.36,success" },
- 55 {
- 56 "type": "output_matches_schema",
- 57 "path": "out/usage-normalized-report.json",
- 58 "schema": { "type": "object", "properties": { "unpriced_models": { "const": { "new-model-x": 1 } }, "total_cost": { "const": 0.36 } } }
- 59 }
- 60 ]
- 61 }
- 62 ]
- 63 }
- 64}