Two plans, no seats and no enterprise tier
These are the planned rates for general release. While ChainOpt is in private beta there is nothing to pay and no card to enter.
Free
planned rate$0/ month
For trying it on one pipeline.
- Up to 1,000 calls analyzed each month
- Redundancy, oversizing and parallelism findings
- Prompt and response evidence on every finding
- Community support
Pro
planned rate$49/ month
For pipelines running in production.
- Unlimited calls analyzed
- Cost history across runs, written to your repo
- CI check that fails a build on new findings
- Support direct from me
What developers ask first
No. ChainOpt runs entirely on your machine. The CLI reads your local traces, computes everything locally, and prints to stdout. Your prompts, responses, and source code never leave your environment, because there is no hosted backend to send them to.
Not from PyPI. ChainOpt is in private beta and the package is not published to any public index, so testers get a signed wheel and the install line together with their invite. Everything after that runs from the chainopt command.
LangChain, LangGraph, and raw OpenAI or Anthropic SDK usage in Python. TypeScript support is on the roadmap.
Only if you ask it to. Analysis is read-only by default, and fixes are suggested as diffs you review and apply yourself.
They are computed from your own observed runs, at real token counts and current API prices, rather than from a benchmark. Findings below the confidence threshold are suppressed rather than guessed. A parallelism finding reports seconds instead of dollars, because running two independent steps at the same time does not lower the bill.
Nothing while ChainOpt is in private beta. After general release the free tier covers 1,000 analyzed calls a month.
Private beta is running now, and I am taking testers.