How Virtum works

Virtum does the work. You stay in charge.

Virtum's seven services hand work to each other, so one request travels from a plan to a running, secured, monitored and documented service. Pick a story to follow it through.

Four stories

Follow the work.

Build an application

For a business that wants an app built and run properly.

  1. Plan it Planner

    Describe what you want. Planner turns it into a plan you can see and change.

    What the assistant does

    • Expands your brief into epics, tasks and acceptance criteria
    • Maps dependencies and the critical path
    • Asks the questions it needs answered

    What you decide

    • Answer the questions and shape the plan
    • Decide which tasks AI agents take and which people take
  2. Build it Builder

    Coding agents work the tasks in Builder, on your own Anthropic or OpenAI account.

    What the assistant does

    • Writes the front end, back end, API and database
    • Works in an isolated sandbox on Virtum's own copy of your code
    • Runs your tests and explains every change

    What you decide

    • Watch it live and steer it in chat
    • Approve each change to a branch and merge request
  3. Secure it Security

    Security checks the code and everything it depends on before it ships.

    What the assistant does

    • Scans code, secrets, dependencies, containers and infrastructure-as-code
    • Ranks findings by real-world exploitation and proposes fixes

    What you decide

    • Approve fixes or accept risks with an owner and expiry
    • Set the gate a release must pass
  4. Deploy it Deploy

    Ship it to your own infrastructure: the clouds, clusters and servers you linked to Deploy.

    What the assistant does

    • Plans environments, containers, DNS and certificates
    • Shows exactly what will change
    • Verifies the release, monitors it and documents it, and rolls back on failure

    What you decide

    • Approve the plan; nothing is applied without it
    • Choose where it runs
  5. Document it Docs

    Docs documents the application, its environments and its runbooks as they are built, and keeps every page current as things move on.

    What the assistant does

    • Writes pages for the app, its environments and every asset it runs on
    • Updates them as the code, infrastructure and configuration change
    • Flags any page that drifts from what is really running, and drafts the fix
    • Answers questions about the system, with sources

    What you decide

    • Review and approve what the AI writes
    • Set owners and review dates, so nothing goes stale unnoticed
  6. Oversee it Operations

    Operations gives you the managerial view once it is live.

    What the assistant does

    • Tracks cost, reliability and incidents
    • Writes the monthly service report

    What you decide

    • See cost, service and risk at a glance
    • Approve requests and changes from one inbox

Run an IT estate

For IT teams and service providers looking after devices, servers and people.

  1. Enrol and watch Deploy

    Every device and server enrolled with a single command, inventoried and monitored.

    What the assistant does

    • Discovers and inventories everything
    • Warns of failures coming, such as disks filling or failing

    What you decide

    • Set monitors, maintenance windows and on-call
  2. Respond Operations

    An alert becomes an incident, a ticket or a task with an owner.

    What the assistant does

    • Correlates alerts and links recent changes as likely causes
    • Proposes the fix and a runbook

    What you decide

    • Approve the fix; it runs on the machine
    • Communicate through the status page
  3. Verify Deploy

    Virtum checks the problem is actually gone.

    What the assistant does

    • Re-checks the device and records the evidence

    What you decide

    • Close the incident with proof
  4. Learn from it Docs

    The postmortem and documentation write themselves.

    What the assistant does

    • Drafts the postmortem and updates the affected runbooks and pages
    • Keeps every asset's page current as the estate changes

    What you decide

    • Review, approve and track the follow-up actions
  5. Serve your people Operations

    Requests, joiners and leavers handled end to end.

    What the assistant does

    • Fulfils catalogue requests and checks they worked
    • Sets up joiners and locks out leavers, then re-reads to confirm

    What you decide

    • Approve what needs approving

Prove your security

For security and compliance leads who need it fixed and evidenced.

  1. Find Security

    Every managed device, repository and cloud account checked.

    What the assistant does

    • Matches advisories to what you actually run
    • Ranks by real-world exploitation

    What you decide

    • Set SLAs and owners
  2. Fix Deploy

    "Patch the critical ones" becomes a staged rollout you sign off.

    What the assistant does

    • Drafts the patch campaign in rings
    • Halts and rolls back if health drops

    What you decide

    • Approve and start the rollout
  3. Verify Security

    Every fix re-checked on the machine.

    What the assistant does

    • Rescans and closes what is fixed
    • Turns anything that did not hold into a tracked task

    What you decide

    • Accept remaining risk with an expiry date
  4. Evidence Docs

    Compliance evidence collected continuously, reported monthly.

    What the assistant does

    • Maps evidence to every framework you follow
    • Builds the monthly and board reports

    What you decide

    • Give auditors their own workspace

Oversee cost and service

For managers who need the whole picture without the detail.

  1. Spend Operations

    Cloud spend, budgets, waste and forecasts on one page.

    What the assistant does

    • Spots anomalies and waste
    • Forecasts month-end and capacity

    What you decide

    • Set budgets and owners
  2. Service Docs

    A monthly service report built from what was actually verified.

    What the assistant does

    • Assembles KPIs, wins, risks and actions

    What you decide

    • Review, approve and share it securely
  3. Delivery Planner

    People and AI agents on one plan, measured the same way.

    What the assistant does

    • Reports AI success, approval and time to commit

    What you decide

    • Set goals and see progress from the work itself

How the services connect

19 handoffs. One system.

Each line is work moving from one service to another without anyone retyping it.

Point at a service or a handoff to see where its work goes.

  • Builder → Planner

    Idea to plan

    An app idea becomes a tracked plan without retyping.

  • Assistant → Approvals → Builder

    Request to code

    "Add a contact form" becomes a change, made only after approval.

  • Planner → Builder → your repository

    Task to merge request

    Ready tasks are built by agents and land only once tests pass and you approve.

  • Builder → Deploy → Operations → Docs

    Release to live service

    Ship it, and it is monitored, given TLS certificates and documented automatically.

  • Deploy → Security → Approvals → Deploy

    Finding to verified fix

    Vulnerabilities fixed on real machines, then re-checked.

  • Security → Planner

    Failed fix to task

    A fix that did not hold becomes someone's job.

  • Assistant → Deploy

    One sentence to a patch rollout

    "Patch the critical CVEs" becomes a staged plan you sign off.

  • Deploy → Operations → Planner

    Alert to owned work

    Alerts become incidents, tickets or tasks, never duplicated.

  • Operations → Builder

    Incident to code fix

    Some faults are diagnosed and fixed in code, with approval.

  • Operations → Docs

    Incident to postmortem

    Every incident leaves a written record.

  • Operations → Deploy, Planner, Docs

    Request to fulfilment

    Staff get software, access or restores, and delivery is checked.

  • Operations → Entra, Google, AWS

    Starter and leaver

    Accounts set up and removed, then read back to confirm.

  • Operations → everything

    Ticket to context

    "What has this server had tickets about?" has an answer.

  • Deploy, Security → Docs

    Work to report

    A monthly report built from evidence, not claims.

  • Deploy → Operations

    Cost to overview

    One page for spend, waste and headroom.

  • Operations → Docs, Planner

    Deal to delivery

    What was sold is exactly what gets delivered and billed.

  • Deploy → Security

    Backup to clean release

    Recovery packages are scanned before anyone receives them.

  • Assistant → Deploy

    AI to your estate

    AI-proposed cloud changes pass the same checks and approvals as any other.

  • Deploy → Academy

    Lessons on real systems

    Practical labs plan restores against your own protected workloads.

How you stay in control

You set the limits. Virtum works within them.

  • You approve what matters

    Every action is risk-rated. Anything beyond the low-risk limits you set waits for a person, and sensitive changes need a second one.

  • The assistant works only inside Virtum

    It reads and changes Virtum's own copy of your code and uses Virtum's tools. Only Virtum's integrations, under your approval, reach your repositories and infrastructure.

  • Your models, your keys

    Use your own Anthropic or OpenAI keys (and a ChatGPT Business or Enterprise plan for coding), or Virtum's own models, with budgets per team and per person.

  • Everything is recorded

    Every plan, change, approval and deployment is on a tamper-evident audit trail you can verify.

See it on your own work.

Design partners run Virtum on real projects and estates, and shape what it does next.