The latest models,
for every team,
within limits
you set.

One workspace for chat, voice and video models from the leading labs. Each job is routed by capability and cost, across only the labs you allow, with spend caps and an audit trail built in.

Small teams start today. Enterprises get a rollout fitted to their organisation.

Your workspace

Routes across models from

  • Anthropic
  • OpenAI
  • Google
  • Mistral
  • ElevenLabs
  • Runway

AI is already in your company. It is just not under anyone's control.

Most companies never chose their AI. It arrived one team at a time, on whatever tool was closest to hand.

  • Subscriptions nobody can see

    Each team pays for its own AI tools on a company card. Nobody has the full list, or the full bill.

  • One model for every job

    The most expensive model writes the meeting notes and the board paper alike, because nobody chose otherwise.

  • Nowhere to point legal

    Legal asks where the data goes and which vendor sees it. Nobody can answer quickly.

Pay for a frontier model only when the job needs one.

The workspace picks a model for each task by capability and cost, from the labs you allow. Admins set the policy once. When a better or cheaper model ships, you swap it in a setting, and nothing your team built changes.

  • Summaries, extraction and translation run on fast, low-cost models.
  • Reasoning and code go to the tier that does them best, with a fallback from a second lab.
  • Every route stays inside the labs an admin has allowed.

Text, voice and video.One login, one bill, one set of rules.

The same permissions, spend limits and audit trail apply whether someone types a question, a voice agent takes a call, or a team generates a clip.

Choose which labs see your work. The others never get a request.

An admin switches each lab on or off, for the whole workspace or for one team. A lab that is off is never used, not as a default and not as a fallback. Bring your own keys and requests go straight to that lab under your own agreement.

  • Allowed labs are enforced on every route: defaults, fallbacks and pinned models alike.
  • Every tier, from quick tasks to video, has a preferred lab and a fallback. Changing one is a single setting.
  • Every change to these settings is written to the audit log.
Admin / Finance labsTry it. Switch a lab off.

Labs

Your key uses no credits
  • AnthropicPlatform key
  • OpenAIYour key
  • GooglePlatform key
  • MistralPlatform key
  • ElevenLabsVoicePlatform key
  • RunwayVideoPlatform key

Defaults

Only labs that are on
TierPreferredFallback
Quick tasksGoogleAnthropic
Deep reasoningAnthropicOpenAI
CodeAnthropicOpenAI
VoiceElevenLabsOpenAI
VideoGoogleOpenAI

Runway switched off by an admin2 days ago

A blocked lab is never used, even as a fallback. Changes are logged.

See every credit, by team and by model.Cap it before it becomes a problem.

Every request is metered in credits and tagged with the team and the kind of model that ran it. Set a cap for the workspace and for each team, and see who is close while there is still time to decide.

Start out of the box.Or have it fitted to your organisation.

Workspace, out of the box

For teams up to a few hundred people. Create a workspace, invite your team and set caps.

  • Create a workspace in minutes
  • Invite your team
  • Set spend caps per team
  • Choose which labs and models are allowed
  • Free plan with 100 credits a month
  • Then pay for what runs

Fitted for the enterprise

For organisations that need the workspace set up around how they work.

  • Your providers, on your own keys
  • Single sign-on
  • Your workflows, built with you
  • A named contact
  • Dedicated support
  • Invoicing

Both run the same workspace. The difference is who sets it up.

Tell us about your organisation. We will show you the workspace on your use cases.

We start with a 30-minute call. Then we set up a workspace on two of your real jobs, so you see the result on your own work. There is no commitment.

Talk to us
We reply within one working day.

Questions, answered.

Where does our data go?

Each workspace is isolated on every read and write. Documents and runs stay in your workspace, access is tied to authenticated identity, and your data is not used to train models. Bring your own model keys and usage goes straight to your provider.

What is a credit?

A credit is a fixed unit of model spend at the platform rate, charged per call after the provider reports usage. Every run shows an estimate before it starts and a live meter while it runs. Runs on your own keys cost no credits.

Which models run the work?

Each job asks for a tier and the workspace routes it to a model from the labs you allow, with fallbacks. Admins set the preferred lab and the fallback for each tier.

Can we control which labs see our data?

Yes. Admins choose which labs are allowed, for the workspace and for each team. A blocked lab is never used, even as a fallback. Bring your own keys and requests go straight to your provider under your own agreement.

What does fitting involve?

We work with the people who own the job to capture rules, voice, examples and exceptions, then configure the workspace to them. Known jobs take days. New ones are scoped as a project.

What if it does not deliver?

Switch the capability off in settings, or tell us and we will. We measure success by work that reaches production, not by pilots.