Your keys. Your agents. Your model.
revoluAItion is plumbing. Every layer is swappable: the worker, the model, the runtime. The Brain stays yours. Everything else is something you can rip out.
Bring your own keys
Plug in Anthropic, OpenAI, Perplexity, Serper. Pay providers directly at cost. No per-seat tax, no markup.
Bring your own agent
Drop in any OpenAI-compatible agent as a worker. It becomes a 5th hand in your Brain, callable from any step.
Bring your own model
Point any step at a self-hosted model. Sovereignty for legal, health, finance. Nothing leaves your tenant unless you allow it.
Featured workers
Use what you already trust. Or whatever ships next week.
Claude Code
Long-context reasoning + strong writing. Default Forge worker.
OpenCode
Self-hostable Claude-Code clone. Good for dev-rel, docs, changelogs.
Hermes-3
Sovereign reasoning. Run it on your own GPU and never leak strategy.
Nemotron
Research-grade. Strong at structured extraction and SERP synthesis.
Ollama
Run llama-3.3 / qwen on your Mac. Zero network egress.
vLLM
Production inference. Bring your finetune, point a step at it.
OpenRouter
One endpoint, 100+ models. Try GPT-5, Grok, DeepSeek without contracts.
Custom
Anything that speaks /v1/chat/completions. If it returns JSON, the Brain will use it.
Connect an agent
OpenAI-compatible? It's in.
If your agent or model serves /v1/chat/completions, point a step at it. The Brain handles prompting, retries, schema validation, and cost tracking.
- Works with any OpenAI SDK shape
- Per-step routing (Forge ≠ Scout)
- Falls back to default on error
- Keys encrypted at rest (Vault)