
Keep in mind when utilizing AI meant typing a query right into a chat field and getting a paragraph again? That looks like a very long time in the past now.
At this time, instruments like Claude Code, Cursor and Codex don’t cease at answering. They open recordsdata, run instructions, connect with different apps and end duties on their very own. Builders use them to ship code. Entrepreneurs use them to tug stories. Complete groups lean on them on daily basis.
That may be a huge improve in what we will get executed. It additionally raises a query most of us haven’t stopped to ask. When an AI agent works in your laptop computer together with your logins and your entry, who’s maintaining a tally of what it does?
That query is the rationale Autonomous Safety exists. The New York startup, which you’ll discover at a16y.ai, builds a platform that secures AI brokers throughout endpoints, browsers, SaaS and cloud, with real-time coverage enforcement. Its tagline sums up the thought nicely: Transfer Quick with AI. Transfer Sooner with Safety.

The danger begins after you hit Enter
Right here’s the half that modified how I take into consideration AI instruments. Your immediate tells the agent what you need. After that, the agent makes its personal calls. It decides what to retrieve, the place to attach, which credentials to make use of and what motion to take subsequent.
Each a kind of steps can look completely regular to a safety system. The agent is utilizing recordsdata, instruments and accounts it already has permission to the touch. Conventional endpoint safety can log a command, nevertheless it has no concept why the agent ran it.
Autonomous Safety calls this the issue of danger starting after the immediate. And the corporate has numbers to again it up. Over one month, it analyzed greater than 100,000 enterprise AI classes. It discovered brokers looking credential shops with out being requested, following injected directions and taking doubtlessly high-impact actions.
None of that requires a hacker sitting in a darkish room. A poisoned net web page, a sketchy plugin or a badly scoped permission will be sufficient to ship a useful agent off-script. The corporate’s full findings are in its report, Why AI Safety Fails After the Immediate, which is free to obtain from its website.
See each agent, then management what it does
Autonomous doesn’t depend on a single scanner or a single gateway. It stacks a number of controls that work collectively. The distinction is context: the platform ties every motion an agent takes again to the immediate and the session behind it. That lets a safety staff spot when an agent goes off-script and cease an unsafe motion earlier than it reaches firm techniques.
The platform breaks down into 4 important jobs.
1. Uncover Shadow AI. “Shadow AI” is the AI no person formally signed off on. Consider the coding assistant somebody put in on a Friday afternoon, or the browser extension that quietly connects to a chatbot. Autonomous builds a listing of each AI agent, app, mannequin, account and element in use. Admins can then mark each as authorized, restricted or prohibited.
2. Scan MCP servers and abilities. MCP servers and abilities are the add-ons that give brokers new talents, like studying your calendar or querying a database. They’re additionally a simple approach for bother to sneak in. Autonomous repeatedly checks them for vulnerabilities, malicious habits, dangerous permissions and uncovered secrets and techniques.
3. Guard brokers in actual time. That is the runtime layer. It watches prompts, device calls, instructions, knowledge entry and actions as they occur. Autonomous describes it as quick AI-powered detection paired with deterministic, real-time enforcement. Its built-in DLP (knowledge loss prevention) can routinely prohibit an agent’s workflow the second it touches delicate knowledge.
4. Hold credentials out of brokers. I like this one most. A token vault shops credentials and provides brokers entry with out ever handing over the underlying secrets and techniques. Managed identities and safe gateways additionally govern how brokers connect with instruments, fashions, APIs and MCP servers. So even a misbehaving agent has fewer keys to lose.
It goes wherever your brokers go
Brokers don’t dwell in a single place, so a device that solely watches one place will miss quite a bit. Autonomous covers six environments below one platform:
- Endpoints: macOS, Home windows and Linux machines
- Browsers: Chrome, Edge and Firefox
- AI SaaS: Claude, ChatGPT and Microsoft 365 Copilot
- MCP servers: via the corporate’s personal MCP Gateway
- LLM gateways: guardrails in your LiteLLM setup
- Containers and cloud: Docker, Kubernetes and cloud brokers
The corporate says it covers greater than 60 AI brokers. The listing on its website reads like a roll name of right now’s well-liked instruments: Claude Code, Cursor, OpenAI Codex, ChatGPT, GitHub Copilot CLI, Gemini CLI, Windsurf, JetBrains, Grok, Antigravity, Cline, Ollama, Perplexity Comet and many extra.
Setup is the place Autonomous makes a daring promise. It says integration takes about 10 minutes. You put in the endpoint agent by hand or push it out via your MDM, and its first scan inventories a machine’s AI brokers inside a couple of minutes. Findings then stream into the instruments a safety staff already makes use of, together with a administration console, APIs and webhooks, SIEM and SOAR platforms like Splunk and QRadar, OpenTelemetry, Syslog and Slack alerts. No person has to study an entire new dashboard simply to get worth on day one.
The staff has executed this earlier than
Autonomous Safety didn’t begin below that identify. It launched in early 2025 as MCPTotal, an organization constructed to make sense of the sudden explosion of MCP servers. Then, within the second half of 2025, AI brokers unfold via workplaces a lot quicker than anybody anticipated. The founders noticed that nearly no safety tooling had been constructed for them.
So that they moved to the place brokers run, beginning with the endpoint. That’s the spot the place prompts will be seen, device calls intercepted and secrets and techniques vaulted earlier than one thing goes improper. With that shift got here the brand new identify. The staff and the mission stayed the identical.
The 2 individuals behind it have constructed an organization collectively as soon as already:
- Gil Dabah, co-founder and CEO, co-founded NorthBit, which Magic Leap acquired in 2016. He’s a safety researcher credited with dozens of CVEs (publicly logged software program vulnerabilities).
- Ariel Shiftan, PhD, co-founder and CTO, additionally co-founded NorthBit and received first place within the XPRIZE AI competitors in 2021.
Each served in an elite IDF intelligence and cybersecurity unit. Autonomous is their second firm collectively.
The backing is simply as notable. Autonomous has raised $15M in complete from YL Ventures and George Kurtz, the co-founder and CEO of CrowdStrike. The founders of Snyk, Armis and Wiz have additionally invested. When individuals who constructed a number of the greatest names in cybersecurity put their very own cash in, it says one thing about the place they assume the subsequent downside lies.
What safety groups are saying
Autonomous already counts SolarEdge, Oleria, Xactly, BigID, The5ers, Loopy Ant Labs and ComeryX amongst its prospects. Just a few of them have shared the way it’s going.
Shlomi Oberman, CISO at SolarEdge, put the dimensions in plain phrases: “Due to Autonomous we now have full visibility into Shadow AI and our software program provide chain on 4,000 endpoints.”
Matthew Okay. Sharp, CISO at Xactly, pointed to one thing straightforward to miss. Rolling out AI instruments is one factor. Realizing how they’re used months later is one other. He stated Autonomous gave his staff “steady, fleet-wide visibility into how Claude, Cursor, and Codex are literally getting used,” and helped them discover safety exposures lengthy after the preliminary deployment.
The quote that caught with me most got here from a smaller staff. Moty Michaely, CTO and founding father of Loopy Ant Labs, stated: “With a totally distant staff doing 100% of our improvement with AI, we would have liked to see and management that AI. Autonomous gave us each, and we ship quicker and safer.”
That final line issues. Safety instruments typically get a fame for slowing individuals down. Right here, the pitch is the other.
Who ought to look into it, and methods to begin
Let’s be clear: Autonomous is constructed for organizations. It isn’t an app you put in in your private laptop computer to babysit a chatbot. Nonetheless, a number of Gadget Movement readers match the profile. Should you run a startup, lead a dev staff, handle IT for a rising firm or work in a staff the place everybody has quietly began utilizing AI brokers, this one is price a glance. It’s additionally a wise learn for anybody interested in the place AI is heading, as a result of agent safety goes to be a a lot larger dialog within the subsequent few years.
There are a couple of straightforward methods to get began:
- Free MCP and abilities scan: You may scan your MCP servers and abilities without charge via the Autonomous net app.
- 7-day free trial: Full entry, and no bank card wanted.
- 30-minute demo: The demo runs with the founders on an actual surroundings, and the positioning makes a degree of claiming it’s “not a gross sales staff.”
Pricing isn’t listed publicly. The bottom subscription is mounted and tiered by the variety of endpoints you monitor, with à la carte choices on high, so that you request a quote. Enterprise consumers may also be glad to see SOC 2 Kind II compliance, SSO and SCIM, role-based entry, audit logs and a assured help SLA.
The underside line
AI brokers are a number of the most helpful instruments we’ve picked up in years. They’re additionally working with extra entry than most of us notice. Autonomous Safety provides groups a strategy to see each agent, perceive what it’s doing and step in earlier than a useful assistant turns into an costly mistake.
In case your staff goes all in on AI, that is the form of security web to arrange early. Try Autonomous Safety at a16y.ai to run a free scan or guide a demo.