Your AI.
Your Hardware.
Your Data.

Trellis is a local-first AI assistant your office runs on its own machine — chat, documents, code, all on your hardware. No cloud subscriptions. No data leaves your building.

The problem with cloud AI

You don't own the AI
you're paying for.

The convenient option became the only option, and now your business runs on a service whose terms, prices, and uptime someone else controls.

  • Rent forever, own nothing.

    Cloud AI is a subscription you're paying indefinitely. Costs scale with use, not with the value you get.

  • Your data, their servers

    Every conversation, every document you ask about, leaves your network the moment you press send.

  • Quotas. Deprecations.

    API rate limits and model EOL announcements set their priorities. Not yours.

The Trellis way

One box.
Your office.
Done.

Local

Trellis runs on your office's hardware. Not somewhere else. Conversations and documents never leave your network.

Predictable

No usage caps. No metered tokens. No forgotten cloud bill doubling overnight because someone left a script running.

Yours

Pick the model. Run it. Replace it whenever a better one comes out. Your stack, your call.

Trellis Server Trainer Clients no cloud
Four surfaces, one assistant

Where you work, Trellis works.

Trellis is one local engine driving four interfaces — built for the actual shapes of work, not a single chat box pretending to do everything.

  • Chat

    Ask questions, draft text, explore ideas. Every conversation saves locally — searchable, persistent, yours to keep.

  • Cowork

    Write documents with the assistant beside you. Brainstorm, revise, restructure — then export the finished result.

  • Code

    Load any source file and ask questions about it. No copy-pasting — the model has full context on what you’ve opened.

  • Train

    Feed Trellis your team’s documents — policies, manuals, internal notes. Every answer searches the index and cites its sources.

How it works

From download to "ask anything," in three steps.

  1. Install Trellis Server

    In your office. One Windows binary, ~30 minutes from download to first chat.

  2. Connect from anywhere

    Desktop, web, or phone — same conversation, every device that’s on your network.

  3. Drop in your docs

    Trellis indexes them locally, searches before every answer, and cites the sources. Your knowledge base, your assistant.

Trust + privacy

We don't have a server.
There's nothing for us to leak.

Trellis runs entirely on your hardware, under your network policy, with the conversation history living in a database you own. We don't operate cloud infrastructure for your data because we don't need to — and that means there's no breach surface where we could betray your trust even if we wanted to.

Stop renting your AI.

Trellis is a one-time setup. Your data stays. Your model stays. Your team gets faster.