Unified LLM spend

One dashboard.
Every model bill.

Last synchronized Waiting for engine… Auto-refreshes every 60 seconds

Combining provider APIs and local events

Current spend $0.00

— imported period

Projected month $0.00

— of $10k budget

Requests 0

0% cache hit rate

Cost / successful task —

— observed success rate

Daily spendAll providers

$500$250$0

Provider splitCurrent period

$0 total

$0Total

Highest-spend projectsWhere the bill goes

Project	Models	Requests	Spend

Top opportunity

Waiting for usage

Connect a provider or record local events to unlock recommendations.

Price opportunity$0

Projects

Spend, attributed.

Every connected vendor normalized into one project ledger.

Project ledger0 active projects

Project	Provider	Models	Tasks	Success	Cost / success	Spend

Efficiency

Cheaper only when
the work still works.

TokenTab compares cost per successful task, retries, latency, tests and quality—without collecting prompts or responses.

Tracked tasks0metadata-only outcomes

Successful—waiting for outcomes

Cost / success—observed spend only

Verified switches0paired local evaluations

Privacy + evidence

Raw prompts, responses, messages and code are rejected. Ten paired tasks produce an observed result; verified switches require 30+ paired tasks and a 95% confidence-bound quality gate.

Connections

One setup.
Separate consent.

Name the private workspace, discover local credentials, then authorize each AI provider independently. No account or email is required.

Zero-account onboarding

Discover your providers

Give this local workspace a name. TokenTab checks the credentials already available to this process without uploading or persisting them.

tokentab — local setup

$ export OPENAI_ADMIN_KEY="…"
$ export ANTHROPIC_ADMIN_KEY="…"
$ tokentab connect openai
$ tokentab connect anthropic
$ tokentab record --provider groq --model llama-custom --cost 0.04
$ npm install https://github.com/adididitagain/tokentab/releases/download/v1.0.0/tokentab-ai-1.0.0.tgz openai
$ tokentab eval --file evals.json --models gpt-5.6-terra,claude-sonnet-5 --allow-provider-data
$ tokentab dashboard

One dashboard.Every model bill.