Horizon, by ACIDRIPP
Infrastructure, run with AI.
Horizon finds every machine, device and service you have, watches them around the clock, and lets you and your AI assistant manage them in plain language. Nothing risky changes without a person's approval.
Request a callOn your own server or hosted by us / from €1 per monitored item per month
What it connects to
- Proxmox VE
- Proxmox Backup Server
- OPNsense firewalls
- TP-Link Omada networks
- Dell iDRAC and Redfish server boards
- AWS accounts
- HAProxy and nginx
- Nginx Proxy Manager
- Cloudflare Tunnel
- Podman containers
- PostgreSQL and Redis
- Linux, FreeBSD and macOS over SSH
- Switches over SNMP
Nothing to install on your machines: Horizon reads them over SSH and through each system's own interface. Product names belong to their owners.
AI management
Most tools added a chat window. Horizon was built so AI can do the work.
Everything Horizon knows and everything it can do is open to AI in a controlled way: it can read all of it, and it can change only what you allow, with a person approving the rest.
01 / Ask
Ask in plain words. The answer comes from live data.
The Assistant is one key away on every page. Ask what needs attention, what happened to a machine or how an incident unfolded. Ask for a change, and it becomes a request that follows your approval rules.
02 / Read
AI reads every log, every hour.
Rules check the logs of all your machines every 15 minutes. Every hour an AI model reads them too and writes down, in plain language, what a person should look at. Serious findings are emailed.
03 / Operate
Your own AI assistant can run the infrastructure.
Horizon has a tool server with 31 tools that speaks the Model Context Protocol. Connect the assistant your team already uses and it can check status, search the inventory, open incidents and ask for changes.
Your AI, your rules
You choose the AI. It can be one that never leaves your building.
Horizon is not tied to one AI company. You pick the model for each job, you can change it in the settings at any time, and a Test button tells you it answers.
AI is a choice, not a requirement. With no model connected, the rules still check your logs every 15 minutes, the Assistant still answers from live status, and everything else on this page works the same.
-
Clouda provider's service
Anthropic, or any service compatible with OpenAI's interface
The quickest start: paste a key, press Test. Keys are kept encrypted in Horizon's credentials vault and are never shown again.
-
Localon your hardware
A model that runs on your own server
Ollama, llama.cpp, vLLM or a command of your own. No key, no outside service: your logs and inventory are read by a model inside your network.
-
Per jobmix them
A different model for each task
The log review, the Assistant and the root-cause engine each get the model you assign: for example a local model for the logs and a larger one for questions.
-
Yoursbring your assistant
The assistant your team already uses
Any assistant that speaks the Model Context Protocol connects with an API key you create, give an expiry date and can revoke at any time.
01 / Discover
Connect a system once. Horizon finds the rest.
Nobody types an inventory. Point Horizon at what you run and it lists what is there, then keeps the list current by itself.
- Connectors for Proxmox, AWS accounts, server management boards and load balancers, set up in a wizard.
- Every host, virtual machine, container and storage pool behind a connector is found and added.
- Network scan and SSH inventory for anything a connector does not cover.
- A network map drawn from what your switches report about their neighbours.
- Manual entries stay yours: discovery never overwrites what a person corrected.
02 / Know
One page for every machine, and it is always right.
What it is, who owns it, what runs on it, what depends on it and everything that ever happened to it. Kept by Horizon, not by a spreadsheet.
- Software and versions installed on each machine.
- Certificates and when they expire, before they expire.
- Relationships and impact: what stops working if this machine does.
- Owners and teams, so every question and approval reaches the right person.
- Capacity, licences and history for each asset, with blueprints that show when a machine drifts from how it should be.
03 / Read the logs
Nobody reads logs. So AI does, every hour.
The warning is almost always in a log somewhere, hours before the outage. Horizon collects the logs of every machine and tells you what matters in them.
- Rules every 15 minutes for the known signs of trouble.
- An AI review every hour that writes a plain-language summary and raises findings.
- Findings with severity, the machine and the lines behind them.
- Resolve, reopen and export so nothing is seen twice or forgotten.
- Email for serious findings, and tamper-evident log archives if you need proof later.
04 / Let your AI operate
Give your AI assistant the keys. Keep the master key.
Your assistant works through the same permissions as a member of staff. It reads freely, and whatever could hurt waits for a person.
- 31 tools: 21 that read (status, inventory, certificates, capacity, findings, history) and 10 that change.
- Changes it can ask for: block or unblock an address or a whole network, update a machine, open and update incidents, resolve findings.
- Approval by default for blocking and for updates. Nobody, human or AI, approves their own request.
- An API key per assistant, with an expiry date, revoked in one press.
- Ready-made tasks for a security review, a health check and triage of a finding.
05 / Watch
The whole infrastructure on one screen, half a minute old at most.
Machines, network, services, hardware and backups, refreshed about every 30 seconds, with alerts that reach the person on call.
- Machines: processor, memory, disk, running processes and history charts.
- Network: firewall, access points and switches, bandwidth per network, live devices.
- Services: containers, web proxies, tunnels, databases.
- Hardware and storage: server health, storage pools and how long until they are full.
- Backups: every job and every machine, so a failed backup is seen the same morning.
06 / Defend
Someone is trying passwords on your servers right now.
Horizon sees it, blocks it on the firewall and tells you. It also sets traps, so someone who is already inside is noticed the moment they look around.
- Password guessing is blocked automatically at the firewall after repeated failures.
- Block a whole network operator when the attacks keep coming from it.
- Decoys: fake files, names and logins nobody should touch. If one is touched, you know at once.
- Sign-in to Horizon with Google and a second factor: authenticator app, SMS or backup codes.
- An encrypted vault for the passwords, keys and tokens Horizon uses. They go in and never come back out.
07 / Change with control
Every change has a name, a reason and an approval.
Updates, firewall rules and everything else that can break something go through one gate. You decide who may do what, and what needs a second person.
- Update machines from the page, with live progress, on Linux, FreeBSD and macOS.
- Roles and permissions down to the single action: allow, refuse or require approval.
- Approvals go to the owner of the machine, with maintenance windows and scheduled changes.
- Automation: flows that react to an alert or a schedule, scheduled jobs and self-heal rules.
- An audit log of every action, and access reviews to check who can still do what.
Horizon in four numbers
All features
Everything in the box.
The full list, in the order of Horizon's own menu. Every customer gets all of it; there are no editions.
01AI8
- AI AssistantAsk about status, a machine's history or an incident, from any page.
- Change requests by askingAsk the Assistant to block something or change something and it files the request for approval.
- AI log reviewAn hourly read of every machine's logs, with a summary and findings.
- Root-cause engineRanks the likely cause of an incident (disk, memory, network, processor) with a confidence score, through the API.
- Tool server for your AI31 tools over the Model Context Protocol, with ready-made tasks.
- AI providersCloud, compatible or local models, assigned per job, with a Test button.
- API keysOne per assistant or script, with expiry and revocation.
- Capability exportA file that tells an AI assistant exactly what it may do in your Horizon.
02Overview3
- DashboardThe main figures, on cards you arrange yourself.
- My viewsYour own dashboards, built from panels.
- SearchOne box for every machine, page and setting.
03Health5
- MachinesLoad, memory, disk, processes, history charts and a timeline for each machine.
- NetworkFirewall, wireless and switches, bandwidth per network, devices online.
- ServicesContainers, web proxies, tunnels, databases.
- HardwareServer health, storage pools and storage runway.
- BackupsJobs, the last good backup of every machine, outside hosts.
04Issues7
- Incidents and problemsQueues, timelines, links to the changes behind them, export.
- SecurityFailed sign-ins, blocked addresses and blocked networks.
- FindingsWhat the rules and the AI found in the logs.
- CanaryDecoys and the record of who touched them.
- Access reviewsRegular checks of who can do what.
- ApprovalsThe rules for what needs a second person.
- Approval inboxWhat is waiting for you.
05Automation5
- FlowsOn an alert or a schedule: notify, open an incident, block an address, call another system.
- AlertingRules, routes, severity and the on-call rota, by email and SMS.
- SchedulesJobs that run at set times, and service level targets.
- Self-healRules that fix a known problem by themselves.
- Change policyMaintenance windows and scheduled changes.
06Resources7
- AssetsThe inventory: every machine, device and service, found automatically.
- CertificatesAll of them, with expiry dates.
- SoftwareWhat is installed where, and in which version.
- LicencesWhat you own and when it runs out.
- RelationshipsWhat depends on what, and the impact of a failure.
- KnowledgeYour team's articles, searchable, linked to the inventory.
- Self-serviceA catalogue of requests your colleagues can make.
07Insights and activity4
- AnalyticsReports, saved templates, service level trends, PDF export.
- CapacityHow much room is left, and for how long.
- Asset historyEvery change to every asset.
- Audit logEvery action taken in Horizon, by a person or an AI.
08Administration and developers8
- People and teamsWho works with Horizon, and in which team.
- OwnersWho is responsible for which part of the infrastructure.
- PermissionsRoles, and your own roles, down to the single action.
- CredentialsThe encrypted vault for passwords, keys and tokens.
- IntegrationsThe connectors and their discovery runs.
- Sign-inGoogle sign-in with a second factor.
- Developer portalA documented, versioned interface for your own scripts.
- Light and darkBoth themes, on desktop and phone.
How you get it
On your server, or on ours.
The same Horizon either way. We install it, connect it to your systems and hand it over working.
- A / Your own server
Installed inside your network
Horizon runs on a server of yours, next to what it manages. Your inventory, logs and passwords never leave your building, and with a local AI model nothing does.
- B / Hosted by us
Run for you in the EU
We run Horizon for you in Frankfurt and connect it to your network over a private, encrypted link. Nothing for you to maintain; new versions arrive by themselves.
Pricing
Two prices, both easy to check.
One for setting Horizon up, one for what it watches each month. These are the starting prices, before VAT; you get a fixed quote after the first call.
-
One time
Infrastructure deployment
from €1,000 one time
We set Horizon up for one infrastructure: installed, connected to your systems and handed over working.
- Installed on your own server, or hosted by us in the EU
- Connected to your hypervisors, servers, network and backup systems
- Automatic discovery fills the inventory on day one
- AI assistant connected to the provider you choose, or to a model that runs on your own hardware
- Sign-in, roles and approval rules set up with you
- Handover session for your team
-
Every month
Monitoring and management
from €1 per monitored item / month
Every machine, device and service Horizon watches and manages for you, billed by the number of items.
- Health, capacity and backup checks around the clock
- AI review of logs and root-cause analysis for incidents
- Security alerts, with blocking of hostile addresses and networks
- Updates and changes with approval and a full record
- New versions of Horizon included
- Support by email
A monitored item is anything Horizon keeps a record of and watches: a server, a virtual machine, a network device, a storage system or a service. The deployment price depends on how many systems are connected and where Horizon runs; hosting by us is quoted with it.
Questions
What people ask us first.
Do we have to install something on every machine?
No. Horizon reads your machines over SSH and talks to your other systems through their own interfaces. It needs an account it may use, which you give it and can take away. Forwarding logs and the decoys are optional extras we set up with you.
Can the AI break something?
It can only do what you allow. An AI assistant works through an API key with the same permissions as a person, the risky actions need approval by default, and nobody approves their own request. Every action is in the audit log with who asked and who approved.
Where does our data go?
If Horizon runs on your server, it stays there. The only thing that can leave is what you send to a cloud AI provider, and you can use a model on your own hardware instead, or no AI at all.
Which AI assistants work with it?
Any assistant or agent that speaks the Model Context Protocol, as the widely used ones do. Horizon's own Assistant is built in and needs nothing else.
We already have a monitoring tool.
Monitoring tells you a number went red. Horizon also knows what the machine is, who owns it, what depends on it and what changed before it went red, and lets you fix it from the same page with a record. You can run Horizon beside what you already have.
Our system is not on your list.
Anything you can reach over SSH works from the first day. For other systems, tell us on the call: connectors are added during deployment and quoted with it.
What counts as a monitored item?
Each thing in the inventory that Horizon watches: a physical server, a virtual machine, a firewall, a switch, an access point, a storage system, a service. We agree the count with you and it is visible in Horizon at any time.
How long does a deployment take?
We agree the date with you after the first call, when we know what you run. Once Horizon is connected, discovery itself takes minutes.
Request a call
See Horizon on your own kind of infrastructure.
Tell us who you are and we will call you. No sales script: we ask what you run and show you how Horizon would handle it.
- We call you within two working days to understand your infrastructure.
- You see Horizon working in a live session, and can ask anything.
- You get a fixed quote for the deployment and the monthly price.
- We deploy it on your server or ours, and hand it over to your team.
No obligation / withdraw the request at any time
Request received
Check your email
We sent a link to . Open it to confirm the request; it works for 7 days. Once you confirm, we call you within two working days.