Utility belt / live
Useful before you create an account.
Each utility solves one bounded problem, explains its data handling, and shows the source date behind model or pricing inputs. The cost calculator anchors the set — all 18 tools listed below are live today; more land one at a time, each with its own tests.
AccountNot required
InputLocal where possible
SourcesDated
Status18 tools live · more planned
The utility queue.
Tools are sequenced by usefulness and data risk. “Planned” means the interface is not pretending to calculate a result yet.
Economics
Cost calculator
Estimate monthly cost across 670+ models from your token volumes — verified prices where sources agree, pending ones clearly labelled, plus a like-for-like model comparison.
Open the cost calculator →Context
Token counter
Count the tokens in any prompt — exact for OpenAI, honestly estimated for Claude and Gemini — with a token map and the same prompt priced across every model.
Open the token counter →Ingestion
OTel validator
Paste an OpenTelemetry span or OTLP trace and check it against the GenAI semantic conventions — required, recommended, deprecated, and mistyped gen_ai.* attributes.
Open the otel validator →Models
Model comparison
Compare 670+ models side by side — price (verified or labelled pending), context window, and capabilities — with a head-to-head for any two, and no hidden missing data.
Open the model comparison →Savings
Caching savings
See how much prompt caching cuts your bill on any model — from your reusable prefix, request volume, and cache hit rate. Shows the cache-write break-even honestly, so a low hit rate that costs more is never hidden.
Open the caching savings →Infrastructure
Self-host breakeven
Find the monthly token volume where running your own GPUs beats paying per-token API pricing — verified API price against your GPU cost, with the crossover point and a throughput reality check.
Open the self-host breakeven →Compliance
EU AI Act classifier
Answer a short decision tree to classify an AI system under the EU AI Act — prohibited, high-risk, limited, or minimal — with the governing Article and Annex cited for every outcome and its obligations. Informational, not legal advice.
Open the eu ai act classifier →Limits
Context window checker
Check whether a prompt and its reserved completion fit any model’s context window — the fit verdict, headroom left, and the input/output limits, from the verified registry.
Open the context window checker →Forecast
Cost spike simulator
Project how your monthly bill moves when traffic scales, prompts grow, or completions get longer — baseline vs projected, the percent change, and which lever drives the spike.
Open the cost spike simulator →Compliance
GDPR readiness checker
Answer a short checklist to see which GDPR obligations apply to your AI/LLM processing and where the gaps are — lawful basis, transfers, DPAs, RoPA, DPIA and more, each cited to its Article. Informational, not legal advice.
Open the gdpr readiness checker →MCP
MCP schema token-cost
MCP tool definitions ride in the model’s context on every request. Paste your server’s tools/list to see the token weight of each tool and the per-request cost overhead — exact for OpenAI, honestly estimated otherwise.
Open the mcp schema token-cost →History
Pricing tracker
Every verified price change on record, diffed straight from the registry’s own version history — cuts and hikes, by model and dimension, dated and sourced. Filter by model or lookback window.
Open the pricing tracker →Ingestion
Prompt hash lookup
Paste a prompt template and see its canonical hash — variable slots detected across mustache, dollar, Python-format, and percent syntax, name-erased, then sha256. The exact algorithm production ingestion uses for prompt_hash, so cost and quality analytics group by template, not by call.
Open the prompt hash lookup →Agents
Agent cost decomposer
Paste a multi-step agent trace and see the verified cost of every step, a per-model breakdown, the top cost drivers, and retry loops burning through the budget — unpriced steps stay visible, never guessed.
Open the agent cost decomposer →RAG
Embedding/RAG cost estimator
Project the cost of embedding your corpus once and searching it every month — initial embed, monthly query cost, and the annual total including re-embeds, on sourced embedding-model pricing verified where sources agree. Vector-DB storage and retrieval compute are not included.
Open the embedding/rag cost estimator →Limits
Rate-limit headroom calculator
Enter your provider’s RPM/TPM/RPD limits and your traffic, and see utilization at average and peak for each constraint, which one binds first, and how many × your traffic can grow before it throttles. Limits are yours to provide — provider tiers change too often to hardcode honestly.
Open the rate-limit headroom calculator →Environment
Carbon/Energy Estimator
Project the energy and CO2e of your monthly query volume as a range across published, cited per-query energy measurements — never a single invented figure. States plainly what the range excludes: training energy, embodied hardware carbon, PUE variance, and non-text modalities.
Open the carbon/energy estimator →Distribution
LLM Pricing API & Widget
A free, keyless REST API over the pricing registry, plus a dependency-free JS widget that drops a model’s live price onto any page — pending and stale prices are always labelled, never guessed. Every embed carries a telemeter.ai backlink.
Open the llm pricing api & widget →