Skip to content
HammerForgeAI

Business · Privacy

It runs where you say it runs.

Storing data in your home country is not the same as protecting it. Two things decide it: where the work runs, and who holds the keys.

Three ways to set it up.

Your building, your own model
  1. Files and recordsKept on your own serversYour building
  2. The searching and the workDone on your own servers
  3. The modelRuns on your own servers too
Web searchOnly if you switch it on

Only this set-up keeps everything in the building.

Your systems, an outside model
  1. Files, databases, recordsStay on your own systemsYour systems
  2. The request and the passages neededGo out to answer it
  3. A model at a providerThe answer comes back

Vulcan runs in your building or your own cloud, and the model is at a provider you approve.

Rented capacity
  1. YouBuy no servers
  2. An agreementBetween you and the provider
  3. The provider’s machinesRun the models for you

The fastest way to start.

  1. The question

    If the company that controls the service is foreign-controlled, a foreign court can order it to hand your data over.

  2. Two laws that reach across borders

    Two countries can reach data held anywhere in the world: the United States and China.

  1. Not tied to one supplier

    We are independent of every AI supplier and every cloud. We recommend the models and the infrastructure that fit your case, and you approve them.

  2. What a build comes with

    The governance documentation your reviewers ask for, so you can explain the system you run and defend the choices in it.

  3. The limit

    On your own machines is a build we do with you, not a switch you flip on a dashboard.

Right-sized

The simplest AI that does the job right.

Not every problem is cold fusion or interstellar travel. We start with the simplest AI that does the job well, which keeps it fast and affordable. We move up only when the work needs it.

  1. Small

    Small models

    • Models that run on a server in your office can do the vast majority of tasks most businesses complete on a regular basis.
    • They run completely privately, on your own server.
    • They are not cost prohibitive.
    • No data center.
    • No sending private information over the internet.
    • QwenOpen source
    • GemmaOpen source
    • Muse GlimmerOpen source
    • Nemotron NanoOpen source
    • GLM-4.7 FlashOpen source
  2. Medium

    Medium models

    • More complex tasks, like computer-aided drafting, complex engineering or analysis, or writing very long documents, are sometimes done better by a larger model than a local server can run.
    • DeepSeek FlashOpen source
    • Nemotron UltraOpen source
    • GLMOpen source
    • GLM-5.3 FlashOpen source
    • Nemotron SuperOpen source
    • Kimi K2.6Open source
  3. Frontier

    Frontier models

    • Solve cold fusion.
    • Analyze the human genome.
    • Drug discovery.
    • Send people to Mars.
    • Write entire code bases with one prompt.
    • GPT AstraProprietary
    • GPT SolProprietary
    • FableProprietary
    • OpusProprietary
    • Kimi K3Open source
    • Qwen MaxOpen source

The sign beside each model shows what it costs to run: green the least, red the most.

Try it on your own work.