Do you see our prompts or data?
No, and not as a policy. The collector runs inside your network and hashes payloads in place. What reaches us is which tool ran, what authority it held, and whether it was reversible.
Pricing
Free where it runs on your machine.
The scanner and the watcher are open source and stay that way. You pay when you want the record kept somewhere other than one laptop, and when you need it to stand up to someone else's scrutiny.
Local
Free · forever
One machine. No account, no telemetry, no expiry.
- Every watch source and every rule
- Authority scan across all five providers
- Usage pulls from your own audit logs
- HTML and JSON reports
- MIT licensed, no dependencies
- Nothing leaves the machine
Team
$12 / agent / month
When agents run on more than one machine, and nobody is going to check a report by hand every morning.
- Everything in Local
- A watchdog that runs continuously, rather than waiting for someone to remember to run a scan
- Slack or email the moment something crosses your threshold, by severity
- Action records collected across machines and CI
- 13-month retention
- Authority drift alerts when an agent's scopes widen
- Unlimited seats; you are billed on agents, not people
- Collector is open source and runs in your network
Evidence
Contact us · annual
When someone outside the company is going to ask what your agent did.
- Everything in Team
- Hash-chained, tamper-evident records
- Seven-year retention
- Underwriter export pack for renewal or a claim
- Public Agent Trust Page for your own sales cycles
- Named contact, annual billing
The rule
We don't charge per finding.
A tool paid by the alert has a financial reason to cry wolf, and the only thing this product really sells is being believed.
On a real machine the first build reported fifteen findings. Eleven were deletions of build and temp directories: all true, none worth reading. Fixing that cut the report to four. Under per-finding pricing that fix would have cost us money, which is exactly why the meter isn't there.
Metered
Agents under watch. That's what the risk scales with.
Not metered
Findings, alerts, seats, scans, rules, reports, retention queries.
What counts as an agent
One deployed agent identity: a support bot, a CI agent, a developer's local assistant. Ten engineers running the same coding agent is ten agents; one agent serving ten thousand customers is one.
Questions
The ones worth answering up front.
What happens if we stop paying?
Local keeps working, because it's the same open-source code. You export your records; we don't hold them hostage to a renewal.
We have two hundred agents.
Per-agent pricing stops making sense somewhere around fifty. Past that it's a flat platform fee. Get in touch and we'll work it out rather than pretending the table scales.
Can we self-host the paid parts?
Yes, on Evidence. The whole design assumes the sensitive half runs on your infrastructure, so hosting the rest there too is a deployment choice, not a rewrite.
Honest status
Local ships today and is what the GitHub repository contains. Team and Evidence are being built. The continuous watchdog, the cross-machine collector and the hash-chained records are not written yet; everything listed under Local is, and you can read it on GitHub today. These prices are what the paid tiers will launch at. If you want either now, you'd be an early customer shaping them, and you should expect to be treated like one.