Poliora helps developers record approved AI usage, compare model costs, and test a lower-cost route before changing a real workflow. Your prompts and source code stay out of the equation.
This route needs Python 3.11 or newer. Don't have it, or not sure? Download the app instead — it includes everything and needs nothing installed first.
Start with Codex or Claude Code usage already on your computer. Poliora shows the models, token totals, plan and quota metadata it can find, before you choose whether to add that metadata to your private dashboard.
$ poliora detect
Reading what your AI tools already recorded locally
Claude Code Max plan 840 requests 190M tokens
Most used model claude-opus-5
6M input · 180M cached · 4M output
Equivalent API value $220.00
Illustrative output, priced with the same catalog Poliora ships: 6M × $5 + 180M × $0.50 + 4M × $25 per 1M tokens. Agent sessions are mostly cache reads, which is why the token count is large and the value is not. Your result comes only from the supported local history on your own computer.
* This is a planning scenario, not a guarantee. Poliora compares the rates and models you approve; validate quality before changing a real workflow.
| Building Phase | Current Route | Candidate Route | Modeled Savings |
|---|---|---|---|
| 1. Product Planning | Premium planning model | Approved lower-cost planning model | Test before rollout |
| 2. UI/UX & Mockups | Premium multimodal model | Approved lower-cost multimodal model | Compare output quality |
| 3. Architecture & Database | High-context reasoning model | Approved lower-cost reasoning model | Benchmark representative tasks |
| 4. Code Implementation | Premium code model | Approved lower-cost code model | Measure production usage |
| 5. Testing & Debugging | Premium debugging model | Approved lower-cost debugging model | Validate before switching |
No. Its launcher scan checks only whether supported tools are available. Poliora records only the usage metadata you explicitly import or connect, never prompts, model replies, source code, chats, or credentials.
Not automatically. Subscription history and provider billing require an authorized export or supported source. A local launcher scan never invents token counts or dollar costs.
The calculator shows a best-case routing scenario, not a guarantee. It applies to usage billed per token. If you are on a flat subscription such as Claude Max or ChatGPT Plus, switching models saves nothing — that fee is already paid, and the lever is right-sizing the plan instead. Actual savings depend on approved models, usage, contract rates, and output-quality testing. Poliora keeps modeled savings separate from proven savings.
Poliora can detect local Codex, Claude Code, Cursor, and Antigravity launchers. It can record approved API usage through supported wrappers and import authorized CSV usage exports.
The dashboard runs on your computer at a local address. Workspace data remains local unless you later choose to export a report or use a future hosted feature.
Your history stays on this computer. The installer opens Poliora when setup finishes.
Heads up: Poliora is not code-signed yet, so Windows SmartScreen will warn you. Choose More info then Run anyway. You can verify the download against the checksum on the release page.
Heads up: Poliora is not notarized yet, so macOS will block the first open. Right-click Poliora in Applications and choose Open, then confirm once.