Virtum
The Virtum platform
Seven services. One assistant, one inbox, one record.
Every Virtum service shares the same foundations: an assistant that works through Virtum's own tools under your approval, one place to approve anything, a tamper-evident audit trail, and models you choose and pay for your way.
The services
Everything you build, run and secure.
-
Builder
Build applications and websites in the browser, by hand, on a visual canvas or with coding agents, all on Virtum's own copy of your code.
-
Deploy
Run everything you own, from laptops and servers to clouds, networks and applications, in one place, with every change approved and recorded.
-
Planner
One plan for everyone doing the work, people and agents alike, tied to the systems it changes.
-
Operations
Run IT as a service, day to day: requests, incidents, people, access and cost, with the evidence attached.
-
Security
Finds what is exploitable in everything you run, gets it fixed, and proves it.
-
Docs
Documentation, runbooks and reports wired to the real systems they describe.
-
Academy
Training built on the environment your team actually runs, with proof of what people can do.
Approvals and autonomy
You decide how far the assistant goes.
Set it per organisation, workspace, project and environment. A lower level can be stricter than the one above it, never looser.
Manual
Ask every time
Everything that changes something waits for a person.
Auto within limits
Low risk runs, the rest asks
Low-risk actions run and are shown to you afterwards; anything riskier waits.
Bypass
Time-limited, visible
For one conversation, it expires, and it is shown while it is on.
Shared by every service
82 platform features.
The assistant One assistant in every service, working through Virtum's own tools.14
- Opens from every page and follows you around
- Conversations you can rename, revisit and delete
- Streaming answers with plan and tool cards
- Attach files and documents
- Takes you to pages, opens records and highlights things
- Knows which service, workspace and page you are on
- Specialist skills in every service
- Personas, effort levels and a model picker
- Long-running sessions you can steer while they run
- Asks clarifying questions when it needs to
- Multi-step plans you approve, run or cancel
- Plan steps run in order; a failed check stops the rest
- Understands "it" and "that one" from what it just did
- A reasoning audit of why it did what it did
Approvals and autonomy You decide how far the assistant goes.9
- Manual, Auto or time-limited Bypass modes
- Destructive actions always ask
- Approve once per conversation, withdrawable
- Approval modes per organisation, workspace, project and environment
- Risk-based rules: confirm from medium or high risk
- Low-risk actions run and are shown afterwards
- A second approver for high-risk actions
- One approvals inbox across every service
- Impact preview before you approve
AI governance and spend Every model call visible, capped and paid for your way.12
- Every model call: who, which model, tokens, cost and outcome
- AI budgets with warnings and hard stops
- Budgets per organisation and per person
- Your own keys: Anthropic, OpenAI, Gemini, Azure OpenAI, Mistral
- Coding agents on your own Anthropic or OpenAI account, or a ChatGPT Business or Enterprise plan
- Choose which providers and models are allowed
- Model catalogue with cost, context and where each model runs
- Automatic redaction of card numbers, IBANs and national IDs
- Prompt-injection detection
- Untrusted documents read in isolation
- AI risk register and AI agent register
- Continuous checks that AI controls hold
Memory and knowledge Context that stays yours.4
- Memory of each conversation, workspace and project
- Long conversations summarised with key facts pinned
- Answers from your docs, runbooks and tasks, with citations
- Export, delete and set retention for your AI data
MCP and extensibility Bring your own tools and agents.4
- A read-only MCP server for your coding agents
- Short-lived, per-session MCP access with the user's permissions
- Connect your own MCP endpoints
- Capability packs and a catalogue of actions the assistant can take
Sign-in and accounts Secure by default.6
- Email and password with two-factor authentication
- Sign in with Microsoft, Google or GitHub
- Your own single sign-on, routed by email domain
- Two-factor required for chosen roles
- Passwords of 12 or more characters, checked against known breaches
- Active sessions you can sign out remotely
Organisations, workspaces and roles Structure that matches your business.7
- Organisations, workspaces and projects
- Belong to several organisations and switch between them
- Built-in and custom roles
- Invitations by email or link
- Modules switched on or off per service
- Ownership coverage and offboarding lists
- Support access only with your approval, time-boxed
Audit and evidence Everything recorded, nothing quietly changed.4
- One audit trail across every service
- Tamper-evident, hash-chained records
- Audit export
- Verification ledger of what is proven and what is contradicted
API and webhooks Build on Virtum.4
- Scoped API credentials with expiry
- Rate limits with headers on every response
- Signed webhook subscriptions
- APIs for the assistant, AI usage, commands and automation
For service providers Run many customers from one place.5
- Linked client organisations
- Pooled billing with wholesale discount and markup
- Per-client statements
- Delegated, time-boxed admin, visible to the client
- White-label branding and custom domains
Data protection and connections Your data, where you want it.6
- Data residency map
- Credential vault with versions, rotation and one-time reveal
- Connectors with metadata-only, snapshot, MCP or on-site modes
- Credential rotation with approval
- One-click AWS, Azure and GCP connection
- Public status page, private until you publish it
Nothing matches that yet. Tell us what you need.
One platform, under your control.
Design partners get early access to the services relevant to them, and shape what comes next.