Vireon DynamicsAI infrastructure studio
Vireon DynamicsLiminal AIFree · open source · MIT
Pricing

Liminal AI is free. Always.

The full agent — every tool, both UIs, MIT license — costs $0. You only pay your AI provider (usually a few cents per task). We charge for optional team support, not software access.

Open Source

Free forever
$0
MIT licensed

Everything Liminal does, on your machine.

  • Full harness — every tool, every feature
  • Bring your own OpenAI-compatible API key
  • Terminal UI + Web UI
  • Persona system, sub-agents, document engine
  • Obsidian vault integration
  • Community support via GitHub issues
  • No telemetry, no account, no middleman
Install now

Team

Coming soon
Contact us
per seat / month

Shared persona profiles and synced memory across your team.

  • Everything in Open Source
  • Shared persona profiles across the team
  • Team-scoped vault sync (Obsidian-compatible)
  • Centralized model + provider configuration
  • Audit log of every tool call (JSONL retention)
  • Private Slack channel with the Vireon team
  • Priority bug fixes

Enterprise

Available now
Custom
annual

Self-hosted, with the Vireon Dynamics team on call.

  • Everything in Team, plus:
  • Custom tool development
  • Air-gapped deployment guidance
  • SSO / SAML integration help
  • Security review + signed evaluation
  • Onboarding workshops for your engineers
  • SLA-backed support, dedicated channel
Cost calculator

How much does the model actually cost?

With two-tier routing and prompt caching on a DeepInfra-backed model, the practical cost per task ends up in the low single-digit cents for typical software work.

~$0.018
Median cost per multi-tool task
DeepSeek V4 Pro · cache hit
13×
Cheaper on rounds 2+
via OpenRouter ephemeral cache
$0
Vireon margin on model spend
Pay your provider directly

Numbers from real session traces, May 2026. Your mileage varies by task length and model choice. The harness logs cache hit rate per turn so you can verify.

Pricing FAQ

Common pricing questions

Is Liminal AI really free?

Yes. Download, install, and use every feature at no cost. The project is MIT licensed on GitHub. We don't sell a Pro tier that unlocks tools.

What do I actually pay for?

API usage to your model provider — OpenRouter, OpenAI, Anthropic, or a local server. Typical coding tasks are often under $0.05 with efficient models.

Does Vireon take a cut of my API bill?

No. Your key talks directly to the provider. We never proxy your traffic or add markup.

Can I use Liminal at my company for free?

Yes. MIT allows commercial use. For teams wanting shared config, audit logs, and dedicated support, contact us for Team or Enterprise plans.

Can I run offline without cloud AI?

Yes. Point the agent at LM Studio, Ollama, or any OpenAI-compatible local endpoint.