See what your AI coding subscription actually gives you.
Tokens, models, effort, quota burn, and price—measured across real coding sessions. Optional check-ins tell us whether the output was worth it.
Measure what your plan actually gives you.
Install once, connect each subscription, and track prompt-free token, model, effort, and quota metadata across coding tools.
Open source, local parsing, and opt-in aggregation. No code or prompts collected. View source ↗
- 01Install the CLI
npm install --global @isaiokay/cliSame CLI via npm, pnpm, or Bun.
Copy to reveal next step
- 02Run onboarding
isaiokaySign in with GitHub, connect detected coding CLIs, and configure the subscriptions you pay for.
Copy to reveal next step
- 03Use your normal CLI
codexKeep using Codex, Claude, Grok, OpenCode, or another harness. Run
isaiokay usagefor the model-and-effort breakdown.
More setup options
isaiokay setup --headlessOpen the link on any device. First-time auth still needs GitHub in a browser.
isaiokay install --allInstall every supported, unconfigured integration.
npx --yes @isaiokay/cli --helpAlso: pnpm dlx @isaiokay/cli --help or bunx @isaiokay/cli --help.
Best coding subscriptions right now
Token allowance, model mix, API-equivalent value, and optional result satisfaction—kept separate and confidence-weighted.
| Rank | Subscription | Value | Observed allowance | Results | Evidence |
|---|---|---|---|---|---|
| — | ChatGPT GoOpenAI · $8/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | ChatGPT PlusOpenAI · $20/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | ChatGPT Pro 20xOpenAI · $200/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | ChatGPT Pro 5xOpenAI · $100/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Claude Max 20xAnthropic · $200/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Claude Max 5xAnthropic · $100/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Claude ProAnthropic · $20/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Cursor ProCursor · $20/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Cursor Pro PlusCursor · $60/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Cursor Start (India)Cursor · ₹649/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Cursor UltraCursor · $200/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Devin MaxCognition · $200/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Devin ProCognition · $20/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | GitHub Copilot MaxGitHub · $100/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | GitHub Copilot ProGitHub · $10/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | GitHub Copilot Pro+GitHub · $39/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Google AI ProGoogle · $20/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Google AI UltraGoogle · Price pending/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Kimi AllegrettoKimi · CN¥199/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Kimi AllegroKimi · CN¥699/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Kimi AndanteKimi · CN¥49/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | Kimi ModeratoKimi · CN¥99/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | OpenCode GoOpenCode · $10/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | SuperGrokxAI · $30/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | SuperGrok HeavyxAI · Price pending/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | SuperGrok LitexAI · Price pending/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
| — | SuperGrok PlusxAI · $100/mo | Pending | Collecting private sample | Not enough contributors | Fewer than 5 |
Subscription value methodology
The factual layer is provider-reported token and quota metadata. Price-normalized value, confidence, and occasional satisfaction are calculated separately so a large raw counter cannot win by itself.
Smallest reliable usage slice
Requests, messages, or turns retain their actual model, reasoning effort, source, and separate token buckets. Mixed sessions are never assigned to one model.
- Exact
- Reported directly by the harness.
- Inferred / estimated
- Shown explicitly and discounted by confidence.
Comparable allowance value
Model-specific tokens are converted using prices valid when the usage occurred, then compared with the plan price. At least 80% price coverage is required.
- Native view
- Input, cache, output, and reasoning tokens.
- Comparable view
API-equivalent value ÷ subscription price
Privacy and confidence gates
Plans stay Pending until five independent contributors opt in. Confidence grows with contributor breadth, complete quota windows, and exact attribution.
- Strongest evidence
- Complete reset-to-reset quota windows.
- Change label
- Observed or possible decrease, never asserted provider intent.
75% allowance value25% satisfactionconfidence shrinkage
Note: Satisfaction is optional and never changes token facts. Effort is part of the model combination—Opus high is not assumed equivalent to Sonnet high. Incomplete windows are lower bounds, not full allowance.