We make money on running the agent, not on building it. Both Managed Agents and Private LLM agents are billed monthly based on actual usage, covering inference, compute, and storage in one line item. Consulting is billed time & materials, hourly, daily, or weekly. Over the lifetime of the engagement, we earn back our build investment, and we're incentivized to keep your agent running efficiently for years.

It's about the deployment, not the model. Managed Agents use a shared, multi-tenant LLM API: fastest to ship, billed by usage. Private LLM runs your model of choice on dedicated, single-tenant infrastructure we provision just for you, whether open-weights (Llama, Mistral, Qwen) or closed-source via private endpoint (Claude through Anthropic enterprise, GPT-4 through Azure OpenAI, Gemini through Vertex AI, etc.). Same cloud convenience, with full data isolation. Essential for regulated industries or sensitive IP.

Yes. You can upgrade any time. We'll provision the private model, port the agent logic over, and run evals so behavior stays consistent. Typical migration takes 1–2 weeks.

Yes. Consulting is billed time & materials, by the hour, day, or week, so you only pay for actual work delivered. Common engagements include architecture reviews, model-selection guidance, eval design, and embedded engineering. A good fit for organizations with established engineering teams who want expert review or additional implementation capacity without handing over the entire build.

You own your agent from day one: prompts, code, evals, fine-tuned weights, and data. We'll export everything at any time. There's no lock-in; we keep clients by being the best option, not by holding code hostage.

Managed Agents and Private LLM plans require a 3-month minimum to cover our build investment. Annual plans get 20% off. Consulting is time & materials with no required minimum; stop or pause any time. After the plan minimum, you can cancel with 30 days' notice. No penalties.