Watching this agent
No issues yetProposed action
AI governance
that runs
on the device.
Oximy runs on Windows and Mac. It checks AI requests, files and agent actions before they go through, then warns, redacts or blocks according to your company’s AI security rules.
This includes bank details
Your policy keeps bank details out of AI. I can remove that column.
50,000+
Devices monitored AI activity across the fleet
6,000+
AI tools tracked Visibility across the AI tools people use
500+
Policy combinations To block activity that crosses your rules
understands the request.
Your policy decides what happens.
Oxi-1 is Oximy’s own model
A new proprietary foundation model, built by Oximy to work out what an AI request or agent action really contains.
- On the deviceRuns locally, only when a request needs it.
- Reads everythingPrompts, files, images, screenshots and spreadsheets.
- Feeds your policyYour rules then allow, warn, redact, ask for review or block.
Oxi-1, our own model, reads the content and context behind each AI request and agent action on the device. Oximy adds deterministic checks, and your company’s AI policy decides the outcome.
The policy engine
What would you like AI to stop doing?
Describe your AI governance policy in plain language. Set what employees and agents can do with AI tools, data and actions, then choose what happens when they cross the line.
Data protection
Keep private details
out of the next request.
Use AI data security rules to check prompts, images, spreadsheets and documents before they reach a supported AI tool.
Coding-agent safety
Stop an agent from
deleting production data.
Choose which files, commands and systems a coding agent can use. Oximy checks each action against your AI agent security rules before it runs.
Clean up the test records.
Destructive actions
Stop deletion of production customer records.
Business controls
Limit the runs. Approve the spend.
Bring AI spend management into your policies. Set run limits, require approval above a budget, and decide which business actions an agent can take.
Routine limits
Set a maximum number of concurrent runs per person.
Usage budgets
Ask before an action exceeds a sample team budget.
Purchase approvals
Review the amount before a subscription is bought.
Mac
It stays on the device.
It shows up when needed.
Oximy brings AI runtime security to supported desktop apps, browser tools and coding agents on Windows and Mac. Checks run locally, and when a rule applies, the employee sees what they can do next.
Desktop AI appsRequests & files
Browser AI toolsAccounts & uploads
Coding agentsCommands & actions
Local device
Oximy policy engine
Deterministic checks + Oxi-1
Click an outcome
Pick an outcome to see what the employee experiences.
The difference
Your policy, applied
before the action.
Oximy enforces AI policy inside the work. It checks files, requests and actions while there is still time to allow, warn, redact, ask for review or block.
When this happensWithout OximyWITH OXIMY
A customer spreadsheet goes into AIThe employee has to spot and remove private details themselves.Bank details are caught in the upload. The employee gets a way to continue without them.
An agent proposes a destructive commandSomeone has to notice which system the command will affect.The production target is checked against the policy. The prohibited command is stopped before it runs.
An automation crosses a business limitThe team has to check usage and spending against a separate set of rules.Your limit becomes a check at the action. A run can wait, or a purchase can require approval.
An employee needs to know what to do nextA policy document leaves them to interpret the rule in the moment.A small notification explains the issue and the next permitted step, inside the work they are already doing.
Backed by 
Advised by security and technology leaders.
Investors
Gokul RajaramProduct leader; board experience at Pinterest and Coinbase
Advisors
Patti Degnan
Former CISO
Notion
Adam Messinger
Former CTO
Twitter
Greg Morrison
Former CIO
Enterprises
Mark Sunday
Former CIO
Oracle
What technology leaders tell us.
They’re tackling problems I know from leading technology at a large enterprise: how technology is used, what value it creates and how to govern it. If I were a CIO today, Oximy would have my attention.
We’ve been fine-tuning the product with Oximy using real data from our environment. Internally, the right people already know Oximy by name. When a reporting need comes up, they ask, ‘Can Oximy do this?’
We’d been looking for a way to see which AI tools people were using, including outside our company accounts. Oximy’s endpoint agent addresses a visibility gap we’d already been trying to solve.
I’ve been championing the product internally and taking it to our legal team, because I see potential our existing tools don’t address. If Oximy delivers, it could become very hard for companies to walk away from.
I know firsthand how hard it is to connect engineering investment to the work people do, and AI adds another layer. Oximy’s focus on visibility makes sense to me.
Understanding AI usage and cost is a useful place to start, and I see a path into the governance capabilities companies will need as they adopt more AI. I’m excited to try it out.
Too much of our picture of AI use comes from anecdotes and self-reporting. What stood out about Oximy is the visibility into the tools people use and the nature of that use.
CIOs are being asked to explain their AI investments to CEOs and boards, and that’s hard to delegate. Oximy is focused on a real issue for those leaders.
I see a promising way to bring the visibility enterprises expect around software usage and spend into AI. That’s highly relevant to the large customers we work with.
Enterprises need a much clearer understanding of AI usage, spend and governance as the number of tools grows. Oximy looks like a solution to a need I expect to keep growing.
FAQs
Before you put it
on your devices.
Oxi-1 is Oximy's proprietary remediation foundation model. It runs locally, on demand, to interpret content and context for security and business policies. It works with the policy engine's deterministic checks for limits, permissions and required approvals.
No. Employees keep working in their existing tools. When a policy needs their attention, Oximy shows a small notification explaining the issue and the permitted next step. The outcome depends on your rule: a warning, redaction, approval request or block.
Yes. Oximy's policy engine and local AI model run on Windows and Mac devices. Policies can cover desktop AI apps, browser AI activity and coding agents; the specific checks available depend on the application and action.
Oximy's content inspection and policy evaluation run locally. A request your policy allows can still be sent to the AI provider you chose. Local inspection and an AI tool's own data handling are separate.
What should AI be allowed
to do in your business?
Bring a security rule, an agent boundary or a usage limit. See how it becomes a policy on your Macs.
Windows macOS