Pricing
Everything that runs on your machine is free and open source (MIT). The cloud is optional: before anything is sent, detected secrets are removed on your machine; personal data is not removed.
zn-gate runs on your machine for free, with no account. An account is only for the optional cloud and Team. Install zn-gate →
Free
$0/mo
No card required
- Calls each month
- 10,000 calls/mo
- Cloud check
- Rules + Galvanize 3
- After the quota
- Prepaid credits, which never expire
- Dashboard
- Personal dashboard and evidence
- Seats
- 1
Pro
$19/mo$190/yr
- Calls each month
- 250,000 calls/mo
- Cloud check
- Rules + Galvanize 3
- After the quota
- $0.10 per 1,000 calls
- Dashboard
- Personal dashboard and evidence
- Seats
- 1
Scale
$99/mo$990/yr
- Calls each month
- 2,000,000 calls/mo
- Cloud check
- Rules + Galvanize 3
- After the quota
- Prepaid credits, which never expire
- Dashboard
- Personal dashboard and evidence
- Seats
- 1
14-day trial
Team
$19/seat/mo$15/seat/mo, billed yearly
14-day trial, card required
- Calls each month
- 50,000 calls per seat each month, shared across your organization
- Cloud check
- Rules + Galvanize 3
- After the quota
- Pay-as-you-go at $0.10 per 1,000, up to your capPrepaid organization credits
- Dashboard
- Team dashboard and exportable audit trail
- Seats
- 2 to 50, with roles
Every call runs the deterministic rules and Galvanize 3: one quota, no separate model count.
Need SSO, SCIM, SIEM streaming or longer retention? Contact us
The Team plan in detail
- Per seat
- $19 per seat per month, or $15 billed annually. From 2 to 50 seats; pending invitations count as seats in use.
- Shared quota
- 50,000 calls per seat each month, pooled for the whole organization. After the pool, pay-as-you-go at $0.10 per 1,000 calls up to a monthly cap the owner sets (by default it stops at the pool); on the annual plan, prepaid organization credits at the same price. Calls count by input size, as on every plan (see “What counts as a call?”).
- Roles
- Owner, admins and members, plus a read-only auditor who can export the audit trail without using a seat.
- What leaves each machine
- Metadata only: secret names, agents, tools, hosts and decisions. Never secret values, prompts or tool output.
- Supported agents
- The shared policy is supported for Claude Code and Codex. Cursor and OpenCode follow in a later release.
More than 50 seats or SSO? Tell us.
Questions
- What counts as a call?
- One request to the analyze API, blocked or allowed, of up to 2,000 characters. Longer inputs count one more call for each extra 1,000 characters or part of them: a 20,000-character input is 19 calls. Every call runs the deterministic rules and Galvanize 3, so there is a single monthly quota. Inputs of more than 2,000 characters are scored in windows across the whole text (up to 20 windows); see the API reference.
- Do packs expire?
- No. Overage pack credits never expire.
- Do you keep my input?
- No. On every plan, inputs are evaluated in memory and discarded after the verdict (Zero Data Retention). Evidence is decision metadata with 90-day retention, visible only to you: it never contains your input, and records from October 12, 2026, 12:24 UTC include a SHA-256 hash of it so you can verify which text a verdict refers to. The one exception is an opt-in data-sharing mode that comes with some startup grants. Paid plans never train on your data. With Team, the audit metadata your machines upload is kept for 180 days.
- How do I start?
- Create an account, generate an API key in the Dashboard, and follow the quickstart.