// SYS.ONLINE — 2501.AI COMMAND INTERFACE

[STATUS: ACTIVE]

2501/Agent specialties

AGENT SPECIALTIES //

WHAT EACH AGENT HANDLES

Agent specialties

Every 2501 agent works within one specialty: what it may investigate, what it may change, and what always stays with a person. Start with your platform to see the problems it handles.

specialtiescategoriesOn-premcloud or hybrid

Find your platform

Pick a category, or browse them all. Each specialty lists the problems it handles and the limits you set.

A Rundeck job chain that stopped partway, leaving later jobs unrun, completes after 2501 fixes the cause and reruns the chain from its start, without adjusting control values to match bad data

Automation and CI

Batch Scheduling

Failed Rundeck job chains, fixed and rerun

Handles

Rundeck job chain stopped partway

Failed CI runs that look alike in a notification are separated by 2501: a download that timed out is retried, a failing test goes to its author with the evidence, and a stuck queue is unblocked

Automation and CI

Failed runs in GitHub Actions, GitLab CI and Jenkins

Handles

Failed CI run: flaky or real?

A Terraform plan that looks like a small update but would replace a production database is flagged by 2501 with what to back up first, and applying it waits for a person

Automation and CI

Terraform Operator

Terraform and OpenTofu plans, drift and state

Handles

Infrastructure drifting from its Terraform configuration

A server ticket for a service that stopped answering is taken by 2501 from diagnosis to a verified fix: the root cause is confirmed before it acts, the service is restarted and the system is checked healthy again

Baseline SysOps

General system operations and incident fixes on Linux servers

Handles

CPU spike above threshold

SysOps Windows starts set to investigate only, so a slow Windows server is diagnosed by 2501 from its event logs and reported with the fix it would make and why nothing was changed, until you raise its level

Baseline SysOps

Windows Server operations over PowerShell and WinRM

Handles

Windows server running slowly

A stopped service, a full disk and access denied that worked yesterday are each traced by 2501 to their cause before anything changes, then fixed with a precise change and tested, within the AWS permissions you grant

Cloud platforms

EC2 incidents and maintenance, investigated and fixed

Handles

Resizing an EC2 instance

An Azure VM that stops answering, a managed disk that filled up and an operation refused for lack of a role are handled by 2501 from a central VM in the subscription, within the Azure role you assign

Cloud platforms

Incidents and maintenance on Azure VMs

Handles

Azure VM unreachable

Compute Engine instances that are unreachable, running a service that is down or out of disk are each fixed by 2501 from a central instance in the project, within the IAM roles you grant, and the result is confirmed

Cloud platforms

Google Cloud Central

Compute Engine incidents and maintenance

Handles

Compute Engine instance unreachable

Failing pods and a service on Kubernetes are investigated by 2501 before anything changes: the pod with a passing glitch is restarted, while a resource limit and a bug in the code are reported for a person with the cause and the change it recommends

Containers and services

Kubernetes Operator

Pod and service failures, handled safely

Handles

Pod stuck in CrashLoopBackOff after being OOMKilled

A systemd service that will not start, a Docker container that keeps restarting and a Podman container that crashed are each investigated by 2501 before anything changes, then restored with the least invasive fix and confirmed healthy

Containers and services

Services Manager

Failing systemd services and Docker or Podman containers

Handles

Systemd service failing to start

A failing service on a Linux host is investigated by 2501 together with what it depends on, and when a restart does not hold it looks deeper instead of restarting again, fixes the real cause and confirms the service stays up

Containers and services

Services Manager Linux

Service failures on Linux hosts

Handles

Service failing on a broken dependency

An IIS site down on a Windows host is restarted on its own by 2501 after it names the cause from the Windows events, so the other sites keep running, and account and security changes wait for a person

Containers and services

Services Manager Windows

Failing Windows services, IIS and containers

Handles

IIS application pool keeps stopping

A red Elasticsearch cluster with an unassigned shard turns green after 2501 finds the reason and retries a recovery that is safe to retry, without deleting the index

Elasticsearch Operator

Red clusters on Elasticsearch and OpenSearch

Handles

Elasticsearch cluster red

A DynamoDB table that is throttled while total capacity looks fine is traced by 2501 to traffic concentrated on a few keys and reported as a data design problem, rather than hidden by raising capacity

MongoDB, Redis, Cassandra and DynamoDB

Handles

Redis running out of memory

Archived logs that filled their disk stop an Oracle Database from accepting changes, until 2501 clears them in step with your backup records, where you allow it, and confirms the database is healthy

Oracle Database DBA work on Unix and Linux

Handles

Oracle archived logs filling the disk

A PostgreSQL service account that reads last month's tables but not yesterday's gets access that covers tables created later, while elevated privileges, configuration changes, restarts and restores wait for a person

PostgreSQL Manager

Roles, access and routine DBA work

Handles

PostgreSQL running out of connections

A SQL Server instance running low on storage gets old backup files cleared by 2501 only after the backup chain is confirmed intact, and the instance is confirmed healthy afterwards

SQL Server Manager

Microsoft SQL Server on Linux and Windows

Handles

SQL Server transaction log full

A new starter, a leaver and a request to add someone to an administrator group are handled by 2501: the new account is created, the leaver is disabled rather than deleted, and the administrator group waits for explicit approval

Identity and directory

Active Directory Operator

Users, groups, Group Policy and DC health

Handles

User cannot log in to the domain

A locked-out user, a switch that drifted from its standard configuration and a failing Wi-Fi login are worked by 2501 through Cisco's own platforms: the user is unlocked, the drift is fixed through your template, and the login problem, caused in the directory, goes to its owner with the findings

Network and edge

Cisco Operators

Catalyst Center, ISE and SD-WAN

Handles

Cisco ISE authentication failure

Requests to open a flow, add an address and bring up a tunnel on a FortiGate are handled by 2501 after it checks what already exists: the new rule respects rule order and is verified, the existing address is reused, and without the VPN key it configures the rest and leaves the key to your team

Network and edge

FortiOS Manager

FortiGate firewall policies, objects and VPNs

Handles

Setting up a site-to-site VPN on a FortiGate

A certificate close to expiry on an HAProxy ALOHA load balancer is replaced by 2501 with its matching key, the change is saved so it survives a reboot, and the applications behind it stay online

Network and edge

HAProxy Aloha Manager

SSL certificates, redirects and load balancing

Handles

SSL certificate expiring on an HAProxy ALOHA load balancer

A ticket to open a flow on a Palo Alto Networks firewall ends with 2501 placing the new rule where the ticket says, previewing the commit with the ticket reference, and stopping if another administrator has changes pending

Network and edge

PaloAlto Manager

PAN-OS security policy, NAT and VPNs

Handles

Adding a security rule on a Palo Alto firewall

An expired certificate, a looping redirect and a site serving the wrong content behind a reverse proxy are each fixed by 2501 with the smallest change and confirmed live, while anything that looks deliberate, such as a maintenance page, stays with a person

Network and edge

Reverse Proxy Manager

SSL, access rules and routing on nginx and Traefik

Handles

Reverse proxy cannot reach the server behind it

A critical vulnerability alert is checked read-only by 2501 host by host: a service open to the internet goes straight to a person, a service reachable only from its host has its severity lowered, and another host turns out not to run the affected version, with nothing patched

Observability and security

Certalert Investigator

CVE alerts on RHEL-based systems, read-only

Handles

Vulnerability alert triage: is it really exploitable here?

Firing alerts are sorted by 2501 into a real problem sent to its owner, an alert that already recovered, a misfiring rule flagged with a correction and missing data, with silences kept narrow

Observability and security

Observability Operator

Alert triage for Prometheus and Grafana

Handles

Noisy alert that keeps firing

On a Linux disk that is filling, 2501 removes old rotated logs past retention, leaves a live log and a database file in place, and reports an application logging errors in a loop as needing a service fix, with how much each would free

Storage and backup

Disk space investigation and safe cleanup on Linux

Handles

Disk partition full

A full LVM logical volume whose volume group has free space is cleaned up safely by 2501, which never extends the volume itself and reports whether growing it is recommended, leaving that decision to a person

Storage and backup

Disk Manager LVM

Safe cleanup for full LVM logical volumes

Handles

LVM logical volume full

A full Windows Server drive is cleaned up by 2501 where Windows space usually goes: update leftovers and temporary files are cleared, event logs are rotated rather than deleted, the most recent memory dump is kept, and shadow copies are left to a person

Storage and backup

Disk Manager Windows

Drive capacity alerts and safe cleanup on Windows Server

Handles

C: drive full on Windows Server

A NetApp volume alert and replication falling behind are investigated by 2501, which finds where the space really went, grows the volume where it has room to grow and explains the replication lag instead of silencing it, never deleting a snapshot that replication depends on

Storage and backup

NetApp ONTAP Operator

Full volumes, snapshots and SnapMirror lag

Handles

NetApp ONTAP volume full

A Veeam Backup and Replication job with a failed and a warning machine is traced by 2501, which retries the temporary failure, escalates the one that keeps failing with its cause, and reports each machine's last usable restore point

Storage and backup

Failed backup jobs and restore points

Handles

Veeam backup job failed

A Proxmox VE node running hot is diagnosed by 2501, which names the guest behind the sustained load and sets out stopping or migrating it for a person to approve

Proxmox VE Operators

VMs, containers and cluster health

Handles

Proxmox VE VM will not start

A vSphere datastore over its threshold gets space back after 2501 removes the snapshots your policy allows and confirms the space returned, while files that look orphaned are reported with their size and kept until their status is confirmed

VMware Datastore Operator

Datastore space, snapshots and orphaned files

Handles

VMware datastore full

A saturated host, a halted VM and a nearly full storage repository on XCP-ng are each traced by 2501 to the right layer, the host's control layer, the host or a guest, the halted VM is started as the ticket asks, and changes to running guests wait for a person

XCP-ng and Xen Orchestra Operators

Hosts and VMs on XCP-ng and XenServer

Handles

XCP-ng host reported busy

How 2501 works

The platform behind every specialty: how tickets reach the agents, how they plug into your tools, and the guardrails around them.

A ticket asks for a fix on a specialty locked to investigate only, and 2501 answers with a diagnosis instead: what it found with the evidence and why it did not act, in a comment on the ticket, with nothing changed

Read-only AI investigation

See how it works

A command from an agent's plan passes every layer that bounds the agent, its access, the blacklist, read-only enforcement, your operational rules and a bounded job, and only the operational rules are an instruction to the agent

Guardrails: what an agent may never touch

See how it works

Universal Gateway

See how it works

Maintenance Automation

See how it works

MCP / A2A plugins

See how it works

Incidence Auto-Resolution

See how it works

Also plugs into

The ticketing, AI models, clouds and operating systems 2501 connects to.

A stopped service, a full drive and a runaway process on Windows Server are investigated and fixed by 2501 over WinRM and PowerShell, optionally signing in with a gMSA so no Windows admin password is stored

Operating systems

Incidents fixed over WinRM and PowerShell, optionally signing in with a gMSA

View technology

An incident, an approved change request and a duplicate arrive in ServiceNow, and 2501 fixes the incident on the machine it names, carries out the change as planned and moves it to review, and resolves the duplicate with the original, writing each outcome back

Incidents and approved change requests, worked end to end

View technology

An agent making mistakes with a complicated command-line tool, the wrong argument or an option that does not exist, instead calls functions an MCP server describes in plain language, assigned only to the agents that need it

Public or in-house tools, given only to the agents that need them

View technology

Repetitive Linux incidents on different hosts, a full partition, a failed service and a CPU spike, are each diagnosed and fixed by 2501 over SSH as a dedicated user rather than root

Operating systems

Everyday incidents fixed over SSH, as an account you create

View technology

The question of which model is approved, rather than which is best, is answered by 2501 agents running on your own Azure OpenAI deployments through Azure API Management, with every command still passing the blacklist

Your own deployments as the agents' models, through Azure API Management

View technology

> READY_TO_DEPLOY

SEE IT ON YOUR OWN INFRASTRUCTURE

Book a technical demo and walk through your own environment with the 2501 team.

REQUEST DEMO

CONTACT SALES