Vector · Agent

The network collector that lives next to your devices

Light, portable and quick to set up: the Vector Agent is one Go program for Linux or Windows. It reaches your local devices over SSH, and sends the results back to Vector over TLS, connecting out by default. In the optional bidirectional mode, which runs through the Vector Hub over VPN, MPLS or private WAN, users can also log in to end devices through the agent with Vector Access. It collects configuration, not performance metrics.

  • Outbound TLS by default
  • Optional bidirectional mode
  • Vector Access
The Vector Agent loopThe Vector Agent pulls a job from the Vector Server, runs it against local devices over SSH, ships the results back over TLS, and a watchdog restarts the agent if it fails. By default the agent starts the connection to the server. In the optional bidirectional mode, the Vector Server can also reach end devices through the agent, via the Vector Hub.CUSTOMER SITEUserUses Vector AccessVector ServerJobs and policyVector AgentRuns at the siteOutbound TLS by default010302Local devicesSSHWatchdog04Restarts01Pull job from serverThe agent asks the Vector Server for work.02Execute against local devicesOver SSH.03Ship results backOver TLS, outbound by default.04Watchdog restarts on failureA watchdog restarts the agent if it stops.Optional bidirectional mode: users use Vector Access through the agent to end devices.The Vector Agent loopThe Vector Agent pulls a job from the Vector Server, runs it against local devices over SSH, ships the results back over TLS, and a watchdog restarts the agent if it fails. By default the agent starts the connection to the server. In the optional bidirectional mode, the Vector Server can also reach end devices through the agent, via the Vector Hub.UserUses Vector Access to end devicesVector ServerJobs and policy0103Outbound TLS by defaultVector AgentAt the site04WatchdogRestarts agent02Local devicesSSH01Pull job from serverThe agent asks the Vector Server for work.02Execute against local devicesOver SSH.03Ship results backOver TLS, outbound by default.04Watchdog restarts on failureA watchdog restarts the agent if it stops.Optional bidirectional mode: users getVector Access to end devices via the agent.

What is a network collector?

A network collector is a lightweight service placed inside a network segment that reaches local devices on behalf of a central platform and reports back. The rConfig Vector Agent is a collector for configuration: it backs up device configurations, runs checks and commands, and sends the results to Vector over TLS, connecting out by default.

The words

Not another monitoring agent

Monitoring tools use the same words, so it is easy to mix them up. A network monitoring probe watches how a device performs. A network monitoring agent reports on the host it runs on. The Vector Agent does neither. It collects configuration, not performance metrics.

  • Network monitoring probe

    Polls devices for performance and availability, so you know if something is up and how it is doing.

  • Network monitoring agent

    Often installed on a host. It reports on that host to a monitoring platform.

  • You are here

    Vector Agent

    A configuration collector. It backs up, checks and runs commands on network devices, and reports configuration back to Vector.

How the Vector Agent compares with direct SSH from a central collector, RANCID and Oxidized, vendor collector proxies and custom cron scripts for distributed network configuration management
CapabilityVector AgentDirect SSH from coreRANCID / OxidizedVendor collector proxyCustom cron scripts
Outbound TLS by defaultYesNoNoPartialNo
Per-site credential isolationYesNoNoPartialNo
Durable local queueYesNoNoPartialNo
Dynamic worker scalingYesNoNoNoNo
SSH, including legacy interactive shellsYesYesPartialSSH onlyPartialVariesPartialVaries
Interactive shell SSH quirks (Avaya, HP, Cisco)YesPartialPartialPartialNo
Windows supportedYesNoNoPartialYes
Structured, size bounded log syncYesNoNoPartialNo
Commercial support availableYesNoNoYesNo

When one collector is enough

If you have one small network, one central collector is fine. If your network has many parts or many sites, or you look after many customers, put an agent in each one. Then you do not have to keep opening firewall rules.

How it works

Get the job, do it, send it back

The agent asks the Vector Server for work. It does the work right next to your devices. Then it sends the result back.

  1. Get a job

    The agent asks the Vector Server for work on a set timer. With the optional bidirectional mode, the Vector Server can also start a job straight away.

    The Vector Agent asks the Vector Server for a job, and in bidirectional mode the server can start oneAt a customer site, the Vector Agent asks the Vector Server for work on a set timer, over TLS, and by default the agent starts that connection. In the optional bidirectional mode, the Vector Server can also start a job straight away, via the Vector Hub.Vector ServerHolds the jobsCUSTOMER SITEVector AgentAsks for workOutbound TLS by defaultOn a set timerOptional bidirectional mode:the Vector Server can start a jobstraight away.The Vector Agent asks the Vector Server for a job, and in bidirectional mode the server can start oneAt a customer site, the Vector Agent asks the Vector Server for work on a set timer, over TLS, and by default the agent starts that connection. In the optional bidirectional mode, the Vector Server can also start a job straight away, via the Vector Hub.Vector ServerHolds the jobsOutbound TLS by defaultCUSTOMER SITEVector AgentAsks for work on a set timerOptional bidirectional mode: the VectorServer can start a job straight away.
    View full size
  2. Do the job at the site

    It runs the job on the local devices over SSH, inside the segment. With the optional bidirectional mode, a user can also log in to a device through the agent with Vector Access.

    The Vector Agent runs the job on local devices, and a user can use Vector Access through itInside the customer site, the Vector Agent runs the job against the local devices over SSH. In the optional bidirectional mode, a user can log in to a device through the agent with Vector Access.UserOpens a terminalCUSTOMER SITEVector AgentRuns the jobLocal devicesSSHOptional: a user logs in to a devicethrough the agent with Vector Access.The Vector Agent runs the job on local devices, and a user can use Vector Access through itInside the customer site, the Vector Agent runs the job against the local devices over SSH. In the optional bidirectional mode, a user can log in to a device through the agent with Vector Access.UserOpens a terminalCUSTOMER SITEVector AgentRuns the jobSSHLocal devicesOptional: a user logs in to a devicethrough the agent with Vector Access.
    View full size
  3. Send back the result

    Results and logs go back to the Vector Server over TLS. By default, the agent starts that connection. With the optional bidirectional mode, the Vector Server can also collect a config in real time, on demand.

    The Vector Agent sends results back, and bidirectional mode adds real-time config collectionThe Vector Agent sends the results and logs back to the Vector Server over TLS, and by default the agent starts that connection. In the optional bidirectional mode, the Vector Server can also collect a config from a device in real time, on demand.Vector ServerKeeps the resultsCUSTOMER SITEVector AgentSends results and logsOutbound TLS by defaultDevicesOptional bidirectional mode:real-time config collection,on demand.The Vector Agent sends results back, and bidirectional mode adds real-time config collectionThe Vector Agent sends the results and logs back to the Vector Server over TLS, and by default the agent starts that connection. In the optional bidirectional mode, the Vector Server can also collect a config from a device in real time, on demand.Vector ServerKeeps the resultsOutbound TLS by defaultCUSTOMER SITEVector AgentSends results and logsOptional bidirectional mode: real-timeconfig collection, on demand.
    View full size
  4. Recover if something fails

    If the agent stops, a watchdog restarts it. Unfinished work replays from the local queue.

    A watchdog restarts the Vector Agent and unfinished work replaysIf the Vector Agent stops, a watchdog restarts it, and any unfinished work replays from the local queue.CUSTOMER SITEVector AgentStopped, then backWatchdogRestarts itLocal queueUnfinished work replaysA watchdog restarts the Vector Agent and unfinished work replaysIf the Vector Agent stops, a watchdog restarts it, and any unfinished work replays from the local queue.CUSTOMER SITEVector AgentStopped, then backWatchdogRestarts itLocal queueUnfinished work replays
    View full size
  5. Nothing stays on site

    No copy of your data is kept at the customer site. Once the results are sent to Vector, the agent deletes them. Device credentials are held in memory only. So the agent is ephemeral.

    Nothing stays on site: results go to Vector and are deleted from the agentAt the customer site the Vector Agent holds device credentials in memory only. It sends its results to the Vector Server, which keeps them, and the agent then deletes its own copy, so no copy of your data is kept at the customer site.CUSTOMER SITEVector AgentCredentials in memory onlyResults and logsDeleted from the agent after sendingSent, then deletedVector ServerKeeps the historyNothing stays on site: results go to Vector and are deleted from the agentAt the customer site the Vector Agent holds device credentials in memory only. It sends its results to the Vector Server, which keeps them, and the agent then deletes its own copy, so no copy of your data is kept at the customer site.Vector ServerKeeps the historySent, then deletedCUSTOMER SITEVector AgentCredentials in memory onlyResults and logsDeleted from the agent after sending
    View full size
Deployment

Put it where your devices are

An agent sits right next to your devices. By default, it connects out to Vector over TLS. Bidirectional mode is optional and needs VPN, MPLS or private WAN. It lets the server reach your devices through the agent. See how the two modes compare.

  1. Customer sites

    Place an agent inside each customer's network. It reaches the devices locally and connects out to Vector.

    A Vector Agent in a customer siteA Vector Agent placed in a customer site reaches the local devices there over SSH, and connects out over TLS by default to the Vector Server.Vector ServerOutbound TLS by defaultCUSTOMER SITEVector AgentInside each customer's networkLocal devicesA Vector Agent in a customer siteA Vector Agent placed in a customer site reaches the local devices there over SSH, and connects out over TLS by default to the Vector Server.Vector ServerOutbound TLS by defaultCUSTOMER SITEVector AgentInside each customer's networkLocal devices
    View full size
  2. Branches and POPs

    Run an agent at the edge, so backups do not have to cross the wide network.

    A Vector Agent in a branch or popA Vector Agent placed in a branch or pop reaches the local devices there over SSH, and connects out over TLS by default to the Vector Server.Vector ServerOutbound TLS by defaultBRANCH OR POPVector AgentAt the edge, next to the devicesLocal devicesA Vector Agent in a branch or popA Vector Agent placed in a branch or pop reaches the local devices there over SSH, and connects out over TLS by default to the Vector Server.Vector ServerOutbound TLS by defaultBRANCH OR POPVector AgentAt the edge, next to the devicesLocal devices
    View full size
  3. Regional data centres

    One agent can look after a whole region. It can also take over heavy polling groups from a central server.

    A Vector Agent in a regional data centreA Vector Agent placed in a regional data centre reaches the local devices there over SSH, and connects out over TLS by default to the Vector Server.Vector ServerOutbound TLS by defaultREGIONAL DATA CENTREVector AgentOne agent for a whole regionLocal devicesA Vector Agent in a regional data centreA Vector Agent placed in a regional data centre reaches the local devices there over SSH, and connects out over TLS by default to the Vector Server.Vector ServerOutbound TLS by defaultREGIONAL DATA CENTREVector AgentOne agent for a whole regionLocal devices
    View full size
  4. Network zones behind firewalls

    Put an agent inside a protected zone. By default it only needs to connect out.

    A Vector Agent in a protected network zoneA Vector Agent placed in a protected network zone reaches the local devices there over SSH, and connects out over TLS by default to the Vector Server.Vector ServerOutbound TLS by defaultPROTECTED NETWORK ZONEVector AgentInside the zone, behind a firewallLocal devicesA Vector Agent in a protected network zoneA Vector Agent placed in a protected network zone reaches the local devices there over SSH, and connects out over TLS by default to the Vector Server.Vector ServerOutbound TLS by defaultPROTECTED NETWORK ZONEVector AgentInside the zone, behind a firewallLocal devices
    View full size
  5. Operational zones

    Run an agent at a remote or industrial site, like a satellite link or a utility plant. It works right next to the devices, so slow or far-away links matter less.

    A Vector Agent in a operational zoneA Vector Agent placed in a operational zone reaches the local devices there over SSH, and connects out over TLS by default to the Vector Server.Vector ServerOutbound TLS by defaultOPERATIONAL ZONEVector AgentSatellite, utility and remote sitesLocal devicesA Vector Agent in a operational zoneA Vector Agent placed in a operational zone reaches the local devices there over SSH, and connects out over TLS by default to the Vector Server.Vector ServerOutbound TLS by defaultOPERATIONAL ZONEVector AgentSatellite, utility and remote sitesLocal devices
    View full size
Security

Safe by design, one agent for each customer

  • Outbound by default

    The agent connects out over TLS with an agent key, so by default you do not need to open inbound ports. Bidirectional mode is optional and needs VPN, MPLS or private WAN.

  • Credentials held in memory only

    Device credentials are delivered per job over TLS and held in memory only. Nothing is stored at rest on the agent.

  • One agent for each customer

    In an MSP set-up there is one agent for each customer, keyed to one rConfig server.

Install

Install in three steps

It is one Go program. It runs as a systemd service on Linux or as a Windows service.

  1. Install the agent

    Install the agent on a Linux or Windows host at the site.

  2. Register it with Vector

    Make an agent key in rConfig and add it to the agent, so it can connect to your Vector Server.

  3. Assign it to a customer

    Tag devices with the agent that should poll them, and assign the agent to its customer.

Vector Agent 2.0

Cold-start memory
18 MBLinux, idle worker pool
Steady-state memory
62 MB8 workers × 25 jobs/min
Binary size
14 MBlinux/amd64, stripped
Avg SSH fetch
241 msCisco IOS show run
Queue recovery
< 2 safter agent restart

Measured on a 4 vCPU / 4 GB Linux collector, 8 workers, 25 jobs/min sustained. Each figure is measured under the condition shown, not a universal constant.

The Vector Suite

Where the agent fits

The agent is one part of the Vector Suite. See the full Vector architecture to learn how the agents, the server and the Central Manager fit together. Running a large, global MSP or enterprise? Our Professional Services team can help you build it with robust support and full control.

Security

Self-Hosted,
Inside Your Perimeter

  • No SaaS dependency

    Every layer runs on infrastructure you control. rConfig has no access to your data unless you grant it.

  • Encrypted in transit

    Agent to server over TLS. Server to Central Manager over AMQP with TLS.

  • Customer separation

    Each customer has its own inventory, jobs and audit history, with hierarchical RBAC across your NOC and scoped access for customers.

  • Built for resilience

    Central tiers support active or warm-standby HA pairs. If an agent stops, a watchdog restarts it and unfinished work replays from the local queue.

rConfig Vector Agent FAQ

What is the rConfig Vector Agent?

The rConfig Vector Agent is a lightweight Go binary that runs as a systemd or Windows service at each site, customer or network segment you need to manage. It connects to local network devices over SSH, and syncs configurations and logs back to a central rConfig server over TLS. By default, the agent connects out to the server.

Is the Vector Agent a network monitoring probe or agent?

No. A network monitoring probe polls devices for performance and availability, and a network monitoring agent usually reports metrics from a host. The Vector Agent is a configuration collector: it backs up, checks and runs commands against network devices and reports configuration back to Vector, not performance metrics. It is designed to sit alongside the monitoring you already run.

Which platforms and protocols does it support?

Linux and Windows. It speaks SSH, including interactive-shell support for legacy devices.

How does the agent work with restrictive firewalls?

By default, the agent only makes outbound TLS connections to the Vector Server. It pulls jobs, pushes results and streams logs on that channel, so you do not need inbound firewall rules. Bidirectional mode is optional. It runs through the Vector Hub and needs a routed path into the customer network, such as VPN, MPLS or private WAN. It adds real-time backups, live interrogation and Vector Access. Opening inbound firewall rules instead is possible, but we do not recommend it.

What happens if the agent or the network goes down?

Jobs and logs persist in a local queue, so nothing is lost. On restart, the agent replays unfinished work. If the rConfig API is unreachable, the agent keeps collecting and queues results locally until it can connect again.

Collect configs at every site, segment and customer

The Vector Agent is one Go program that runs as a systemd or Windows service. Put it anywhere your devices are, and let rConfig do the rest.