Its own computer
Each agent gets a real machine. Watch it work, or take control any time.
On-premise AI agents that get work done for you
Miles agents work alongside your team around the clock. Run them in our cloud, your AWS, Azure or GCP account, or fully air-gapped.
Archieneeds your OK×
Refund the 37 customers who were charged twice?
- Amount
- €1,284.50
- Through
- Stripe refunds
Works in the apps your team already uses
Chatbots answer. Agents act. A Miles agent works on its own computer, in the apps your people use, and asks before anything that matters.
Asks before it acts
Payments, outside email and deletes wait for a person.
Every action on the record
Each step lands in a hash-linked log no one can edit, including us.
Run AI agents on your
own infrastructure
Agents need deep access to your data. Miles brings the agents to the data, not the data to the agents.
- Miles Managed CloudWe run it for you.
- Your CloudAWS, Azure or GCP.
- Your PremisesOn-premise, including air-gapped.
Features
Three places, one screen
Each says who runs it, where data lives and what leaves your network.
Bring your own cloud
AWS, Azure or GCP, in the region you pick, with a guided setup and a connectivity test.
Models on your premises
Run the model inside your network, so prompts never leave it. Choose which roles may use each.
Kill switch
One typed phrase stops every agent and sleeps every computer, wherever it runs.
Who pays, stated
Each computer shows who gets the bill: plan hours, your cloud, or your own hardware.
AI agents for
every team
Each team gets agents for the slow, repeatable part of the job. People make the calls.
Sales
Leads, ready by morning
Researches and enriches leads overnight, and notes why each one is worth a call.
Engineering
Regressions caught before review
Writes and runs extra tests on every pull request, and flags what broke with a failing test attached.
Ops and execs
Calendars and dashboards
Rearranges the day around what matters, and builds dashboards from your internal data.
Research and strategy
A sourced brief
Reads your documents and the web, and returns a brief with every claim tied to its source.
Creative
References for a scene
Generates frames and gathers references for a scene, saved where your team can review them.
Hand it
the work
Describe the task in chat. The agent does it on its own computer, in your apps, and comes back with the result.
Learn more about AI agents with their own computer#financeSlack
Priyanshu10:02
Can someone reconcile the October vendor invoices before Friday?
Santosh10:04
Most are in Gmail, the rest in the vendor portal.
@Ledger reconcile them and flag anything that doesn’t match
Features
Live computer
See the agent’s screen in the chat while it works, or pop it out.
Take control
Step in for a sign-in or a code, then hand it back.
Steer or queue
Type while it works: steer changes the task at its next step, queue waits until it’s done.
Continue where it stopped
If a task fails, its progress is saved. Fix the blocker and the agent picks up from there.
Asks before
it acts
Anything that sends, pays or deletes stops for a person, with the exact action and amount in front of them.
Learn more about approvals for AI agentsArchiewants to refund 37 double charges
- Action
stripe.refunds.create- Amount
- €1,284.50
- Approvers
- You, then Priyanshu
Features
Hold to approve
Risky actions need a deliberate press and hold, never a stray click.
Second approver
Money above a threshold needs a second person, never the one who asked.
Standing grants
Let an agent repeat a safe action up to a limit and a date. Every use is counted.
Admin inbox
Approvals, budget requests and app requests in one place.
Every action,
on the record
Who did what and from where, with the value before and after. Signed every hour, so the log proves nothing changed.
Learn more about the audit log
Features
Person or agent
Every event names its actor, so you can see what an agent did for whom.
Integrity check
Hash-chained digests every hour. Verify the log at any time.
Filters you can share
Filter by category, person and time, and send the exact view as a link.
Export
Download what the filters show as CSV or JSON. The export is on the record too.
Enterprise AI agents,
run in one place
Roles, budgets, apps and sign-in for the whole company, with spend by person and a forecast against the budget.
Learn more about running AI agents across your company
Features
People and roles
Owner, admin, billing admin and member, each in plain words.
Budgets
A budget per person and for the workspace, with alerts as it fills.
SSO and SCIM
Connect your identity provider, test sign-in, then require it.
Apps policy
Allow or block apps, see who connected what, and revoke one account or all of them.
Every employee, several agents. That’s the future of work we’re building for: people set the goals and make the calls, and their agents work around the clock, inside your own boundary.
Questions about
on-premise AI agents
Anything else, ask us on a call.
What is Miles?
Miles is enterprise software for running AI agents on your own infrastructure. Each agent works on its own computer, alongside your employees, around the clock.
How is an AI agent different from a chatbot?
A chatbot answers questions. An agent does the work: it signs in to your apps, finishes the task, and asks before anything that sends, pays or deletes.
Can Miles run on-premises?
Yes. Run agents on your own Macs and Linux hosts, or all of Miles inside your network, air-gapped. You can also use Miles Managed Cloud or your own AWS, Azure or GCP account.
Does our data leave our network?
Not when Miles runs in your cloud or on your premises. Computers and files stay inside your boundary; only status, usage, the live screen while someone watches and what the agent sends to its model leave it. Air-gapped, nothing does.
What can an agent do without asking?
You decide, per kind of action. By default, sending, paying and deleting wait for a person, and money over your threshold needs a second approver.
Which apps do Miles agents work in?
Slack, Gmail, GitHub, Notion, HubSpot, Salesforce, Stripe, Zendesk and more. Admins choose which apps agents may use, and the most they can do in each.
Open this week
Put agents to work inside your own boundary.
Bring a real task. We’ll show an agent doing it on its own computer, in your apps, and asking before it acts.
+3
Book a call