Plans & pricing

Ironhide is priced on usage: how many agents you protect and how many runs you execute each month. Every tier gets the same engine and the same results, nobody gets a weaker check for paying less. Bigger tiers get more protected agents, more monthly runs, and heavier tooling around the result.

Data rights are not a pricing axis and not a feature you toggle. Every run, on every tier, is owned — there is no privacy tier to buy and no per-agent opt-out. What Ironhide keeps and reuses is the scrubbed environment your agent ran against, never your agent's own trajectory; the mechanics are on the data rights page.

The four tiers#

Tiers are cumulative: each includes everything below it. Items marked coming soon are on the roadmap and not yet available.

What counts as a run. One run is one CI test session: one ironhide test, however many tests and samples it drives, and however many it drives at once. A nightly run is one run. A pull request whose code did not change since its last graded run reuses that result and is not counted. Runs are counted per calendar month (UTC). Enterprise has no self-serve checkout — email hello@ironhideai.com.

The Free tier#

Every new account starts on Free: 1 protected agent and 100 runs a month, free, no card. That is enough to wire the gate into a PR and see real results before paying anything. Two things to know:

• Every run is owned-data. This is true on every tier, free and paid alike: the shared library is built from the scrubbed environments your agent runs against, not from your trajectories, and there is no opt-out to manage (data rights).
• Quotas are enforced by the server, not the honor system. The run meter counts against your tier's monthly quota, and the agent cap counts registered agents. When a quota is reached, new runs are refused with a clear message that names the quota and the tier, never reported as a pass; your agents stay registered and every past report stays readable. The built-in demo agent is exempt from the meter.

The catalog API#

GET /plans returns the live catalog — the same four tiers this page describes, cheapest first: developer, starter, pro, enterprise. Plan ids are not display names: developer is the Free tier and starter is the $29 Developer tier — use the id when calling the API (for example ironhide upgrade --plan starter). Each entry carries the tier's id, name, price_monthly (dollars; 0 for Free, null for Enterprise's contact-only pricing), agent_quota and run_quota (null = unlimited), and the tier's feature entitlements. The web UI and this page show the same catalog; if they ever disagree, GET /plans is the truth.

Choosing and changing plans#

• GET /plans lists the catalog.
• Selecting a paid plan routes through checkout (POST /account/checkout with {"plan_id": "..."}). Selecting a plan grants nothing until the subscription is active — until checkout completes you are metered on the Free quotas, so a half-finished upgrade can never grant paid entitlements.
• Upgrades go through checkout at the higher price; the price difference is stated up front.
• Downgrades apply against the lower tier's agent and run quotas; if you are over the target tier's caps, reduce your agent count first.

Why the numbers look like this#

• Free is a real tier, not a trial. 100 runs a month is enough to gate a small project's PRs indefinitely. The front door is a developer seeing a real result in CI, not a countdown.
• $29 is a personal-card decision. One developer can turn it on for a side project or a second agent without asking anyone.
• $99 is a team's tool budget. Ten agents and 3,000 runs cover a team running the gate on every PR, plus the compliance export for the security review.
• Each step up is the better deal. Every tier gives more agents and runs per dollar than stacking the one below it — five Developer plans cost more than Pro and give less.
• Data rights are never the upsell. Every run is owned on every tier. What you pay for as you go up is capacity and tooling, never a change to how your data is handled.