Free to run. Paid when a change needs judgment.

Finding the problems costs nothing, and neither does correcting the ones with a single right answer. You pay when a change needs judgment, and for the team machinery around it. Below: the four tiers, what the free one covers, common questions from procurement, and common objections.

Four tiers. Each one adds the work the tier below cannot finish alone.

Tiers are cumulative and enforced in the product, so nothing you rely on disappears when you move up.

Per seat · change the count any time · keys validate offline

Free

$0
forever · Apache 2.0 engine
Includes
The analysis engine with core-language and linter checks
Supported mechanical fixes, applied and verified
MCP server, SARIF and CI output
No account, no API key
Install it

Pro

$39
/ seat · month · change any time
Everything in Free, plus
Cataloged refactors, one function at a time under a constrained prompt
PDF audit reports you can hand to someone
Delta scoring between runs
UX audit
Component decomposition

Max

$89
/ seat · month · change any time
Everything in Pro, plus
Structural rewrites, with the evidence behind them
Custom rules for your own conventions
Cross-file analysis
Discovery and planning

Teams

Talk to us
a support layer on Max, with no fourth price
Everything in Max, plus
A person to reach, first
Procurement and licensing across the org
Onboarding for the team
Contact us
Pro covers what these three do separately. Max sits at roughly twice Pro, where Snyk puts its upper tier.
ToolPriceCovers
CodeRabbit$24–$30 / dev / moAI code review
Codacy$18–$21 / dev / moStatic analysis
Snyk$25 / dev / moSecurity scanning

Checked against each vendor’s public pricing page, September 2026.

What the free tier covers.

37.1% of the patterns carry a mechanical correction with one right answer. Those are applied and verified for nothing, on the free tier, with no account and no key. You can prove that on one repository this afternoon before any money is discussed.

A licence buys the other two tiers, where something is proposed and a person signs it off. Nothing is metered and nothing expires.

Applied for you
37.1%
Bounded refactor
57.7%
Your call
5%

Shares count patterns in the library, which is a different count from your findings. What each tier does

Terminal, free tier
$ pip install truemend
$ truemend .
Health 6.6 · 47 files · 312 issues
115 of them correct themselves for $0
$ truemend fix . --apply

The six questions you will be asked before this gets approved.

Written so you can forward this section as it is.

  • How do we pay?

    You create an account, then check out through Stripe. Card details, invoices and cancellation live on Stripe. This site keeps your customer ID, tier and seat count, and no card data. The plan goes live when Stripe confirms it.

  • Where does the licence key come from?

    You generate it yourself in the portal once the plan is active. Signing writes your organisation name, tier, seat count and a one-year term into an RSA-2048 key. Generating a new one marks the previous one revoked in your account, so exactly one is live there. A key already saved on a machine keeps validating offline until it expires.

  • Does it phone home?

    No. The public key is embedded in the binary, so a signed key validates locally and works air-gapped. Issuing, verifying and revoking happen on our side, in the portal. The check on your machine does not reach us.

  • What counts as a seat?

    A seat is the quantity on your subscription, and you change it in the portal: billing follows on the next invoice. Team members you invite inherit the plan without a subscription of their own.

  • Can we stay on free?

    Yes, and plenty do. Detection and every mechanical correction, applied and verified. Nothing expires and nothing nags you. The free tier needs no account at all.

  • What is it licensed under?

    The engine is licensed under Apache 2.0. Detection content is licensed under Elastic License 2.0. A paid plan adds checks and seats; it does not change either licence.

  • Or hand us the repository.

    If you would rather not run the first pass yourself, we do it: a scored report, the supported fixes already applied, and a walkthrough with whoever needs convincing.

    How the engagements work →

Licences, seats and keys all live in the account portal. Setup questions go to support.

Four common objections.

  • We already have agents that clean this up.

    In-loop cleanup is usually a model checking its own output, with no published accuracy figure. Truemend finds problems with syntax-tree analysis and no model, across the whole repository, and publishes its precision. It also runs as an MCP server and a CI check, so your agents can read its findings.

  • Isn’t this Semgrep or SonarQube with a new name?

    It’s the same category. The differences are the published comparisons against those tools, losses included, and the repair step: findings arrive sorted into three tiers, with the mechanical ones already fixed.

    See the results →
  • Won’t it flood us with false positives?

    Only the first tier, 37.1% of the library, changes code by itself, and only where the fix is mechanical and the file re-parses. Everything else comes back for a person to accept or reject. You can switch off any pattern per repository.

  • Why pay when our IDE agent does this for free?

    For detection and mechanical fixes, you don’t need to: both are in the free tier. Paid plans cover changes that need judgment, and team features such as seats.

Tell us about the codebase.

Teams adds a person to reach, procurement, and licensing across the org on top of Max. You get back a real scope, from someone who has read what you sent.

Only a name, an email and your company are needed. The rest lets us come back with something specific instead of a call to discover it.

Goes to hello@truemend.ai