We are building the control plane for agent teams.

ALDO AI is a product of ALDO TECH LABS, a software studio obsessed with one question: what does it look like when a real engineering team is staffed mostly by LLM agents, and an actual control plane sits between them and your code? Most "agent frameworks" today answer half of that question and stop. We wanted to ship the other half.

Why we built it

The current generation of agent stacks bakes a single provider into the framework, treats local models as a second-class afterthought, and leaves privacy enforcement to whoever is writing the agent prompt. That was never going to work for the kind of team we wanted to run. So we drew a line at the first commit:

• LLM-agnostic by construction. No code path names a provider. Agents declare capabilities; the gateway picks a model. Switching providers is a config change, never a code change.
• Local models are first-class. Ollama, vLLM, llama.cpp, LM Studio, MLX — auto-discovered, evaluated against frontier models on the same agent spec.
• Privacy tiers are enforced by the platform, not the prompt. An agent marked sensitive is physically incapable of reaching a cloud model. The router fails closed.
• Every run is replayable. Full message and tool-call history; re-execute any step against a different model.

Different from "DIY your own framework"

You can wire most of this together yourself with enough glue code. We have done it. The cost of that path is: every privacy invariant becomes a code review; every model swap is a refactor; every "what did this run actually cost?" becomes a SQL query. ALDO AI is the assertion that those things should be platform features, not engineering folklore.

The reference agency

Every new tenant is seeded with the same agency/ folder we use internally — principal, architect, engineers, reviewers. It is the unit test for every platform feature: if a feature does not help that agency ship, we do not ship it.

License

ALDO AI is proprietary. Access is granted to paying customers and design partners under our standard terms of service. The hosted product, the SDKs, and the integration surfaces are all available through the trial — see Pricing.

Get involved

The fastest way to talk to us is the design-partner program — a small group of teams getting hands-on access while we shake out the rough edges. Or just drop us a line.