Pricing

Undo, stop and the rules are free.

One paid plan. It buys the phone remote and an outside party's signature on the time.

Free
$0
MIT, forever, no account
  • Undo with pr undo, and pr stop and pr resume for every agent that runs the hook on the machine
  • The loop breaker and file lanes
  • pr scan for a repository you just cloned, and pr brief for reviewing what an agent changed
  • The rules, 45 armed by default, and day, session and total spend caps enforced at the tool boundary (pr guard budget 25)
  • The same rules in an interactive session, in claude -p and in CI
  • /guard-status, /guard-card and the local decision journal
  • Signed, hash-chained records (pr guard receipt) and both open-source verifiers
  • pr spend, and pr reconcile against a provider invoice
  • One independent RFC 3161 timestamp, free, so you can see what an outside party's receipt looks like on your own run
  • Community support
Install the plugin

Nothing that undoes a change, stops an agent or refuses a tool call is behind a payment, and nothing is metered. The licence check on the remote is a commercial control in open-source code, not DRM. Your agent records stay on your own infrastructure either way, including when you pay. 14-day refund.

What changes when you pay

Little, and that is deliberate. Undo, stop, the engine that refuses a tool call, the cap that stops the spend and the cryptography that makes a decision checkable are the same code whether or not you pay. Money buys two things: a way to answer an agent from your phone, and an outside party's signature on the time, which you cannot produce for yourself.

Scroll the table sideways if it does not fit.

CapabilityFreePaid plan
Undo: checkpoints before every action that can change a filepr undo, git repositories onlyYesYes
Stop and resume every agent that runs the hookYesYes
Loop breaker and file lanesYesYes
pr scan and pr briefYesYes
Guardrail rules at the tool boundaryYesYes
Spend cap that refuses the next tool callday, session or totalYesYes
The same policy in claude -p and CIYesYes
Signed, hash-chained decision receiptsYesYes
Both open-source verifiersYesYes
Reports and self-contained packs from the CLIYesYes
AI code attribution (pr attest)signed, local, free for goodYesYes
Phone remoteTelegram or ntfy, no Provenrail server in the path14-day evaluationYes
Anchoring the root of your chainto our service, or to a server you runOne, freeUnlimited
RFC 3161 timestamp from a public authorityOne, freeUnlimited
Public receipt page an outside party can openOne, freeUnlimited
Run the whole thing yourself insteadSelf-host the AGPL serverSelf-host the AGPL server

Pricing questions

Is anything metered?
No. There is no quota to run out of. Nothing is metered by usage. The policy engine runs on your machine, it makes no network call to decide anything, and the paid option is a flat monthly licence rather than usage.
Is the spend cap a paid feature?
No. It is MIT licensed and free in every released version. Charging for the thing that stops a runaway bill would be a strange way to run this. pr guard budget 25 works on a free install with no account.
Does the free option expire?
No. It is MIT licensed source you already have. There is no trial clock and no card required for anything in it. The one thing with a clock is the phone remote, which runs as a 14-day evaluation that needs no account and then needs the paid plan. The 14 days are counted on your machine from the day you run pr remote setup. When they end, questions stay on the machine exactly as they did before you set the remote up, and nothing else changes. No card or account was taken, so nothing is charged; the remote starts again only if you buy the plan and run pr activate.
Is undo a paid feature?
No. It is MIT licensed and free in every released version. pr undo, pr stop and pr resume, the loop breaker and file lanes work on a free install with no account. Charging to get your own files back would be a strange way to run this.
What does the phone remote do, and what leaves my machine?
It carries an agent's question to Telegram (a bot you create) or ntfy (a secret topic) and the answer back, and it can stop or resume every agent. Provenrail runs no server in that path. The text of a question, meaning the rule id, the tool name, the project folder name and by default up to 600 characters of the command, passes through the service you chose, under that service's terms. The privacy page has the exact list.
Is the licence check DRM?
No. It is a commercial control in open-source code: you can read it, and the evaluation and the paid plan are how the remote is sold. Nothing that undoes a change, stops an agent or refuses a tool call depends on it.
Is the cryptography weaker without paying?
No. Hash-chain integrity, Ed25519 signatures and the open-source verifier are the same code either way. Paying adds RFC 3161 timestamps from an external time authority, which prevents back-dating; it does not make the chain itself any stronger.
Do you store my agent records?
No. Provenrail hosts identity and billing only. Agent records live in your own sink, on your own infrastructure. That is a deliberate design choice, and it is why the verifier runs entirely in your browser at provenrail.com/verify with no upload.
Can I self-host instead of paying?
Yes. The client, SDK, verifier and spec are MIT-licensed and the server is AGPL-3.0-or-later. Run the whole thing yourself at no cost. The paid option exists for external time and independent anchoring, not to unlock the core. There is no hosted tier to buy: you run the sink either way.
What is independent anchoring, and what leaves my machine?
Your records are hashed into a chain. Independent anchoring sends us the single root of that chain, 32 bytes, plus how many records it covers. Nothing else is sent, because the endpoint has no field anything else could arrive in. We timestamp that root, keep an append-only history of it, and publish it at a URL anyone can read without an account. Your client or auditor recomputes the root from the records you show them and checks it matches. If you later edit or delete a record, it stops matching, and neither you nor we can make it match again.
Why can I not just anchor my own chain?
You can, and the tooling is free. What you cannot do is be your own independent party. A timestamp you produced yourself proves the chain is internally consistent, not that it existed before the dispute. That is the only thing a paid plan sells that self-hosting cannot reproduce, which is also why the first one is free: you should see the difference on your own run rather than take this paragraph's word for it.
Is the hosted anchor service running today?
Yes. pr anchor-push bundle.json sends a 32-byte fingerprint of your records and nothing else, and returns a receipt carrying an RFC 3161 timestamp from a public authority. There is no signup step beyond buying it: the licence key you activated is the credential, and the command picks it up on its own. You keep every record; we never receive one, which is why this is the one part of Provenrail we can host. Every account gets one anchor free, so you can see the receipt before you decide.
Why do you charge for the remote and for anchoring?
Anchoring has to run on someone else's machine to mean anything: a timestamp you issue yourself proves your chain is internally consistent, not that it existed before the dispute. The remote is the convenience of answering from your phone. Everything that undoes a change, stops an agent or refuses a call, the rules, the cap, the journal, the records and the verifiers, runs entirely on your box, so it is free.
What can you do with what I send you?
Almost nothing, by design. An anchor request carries a stream label, a 64-character fingerprint of your records, and how many records it covers. There is no field a record could arrive in. We cannot read your prompts, reconstruct your chain, or show anyone what your agents did, because we were never sent it. What we can do is refuse to sign a rewritten history: coverage of a stream only grows, and the same length cannot get two different fingerprints.
How do refunds work?
A full refund within 14 days of a purchase: email [email protected] and Polar, the merchant of record, processes it. The subscription renews each month until you cancel, which you do from the billing portal on your account page.
Can I pay annually, or in euros?
Billing is monthly in US dollars today. If annual billing or another currency is a blocker for you, write to [email protected] and say so; that is the kind of request that decides what ships next.
← Back to home