Patrick Van Deven, CEO and founder
Since we founded VaultSpeed in 2018, our core design principle has been to work top down from the data model, because that is where the context lives, and to automate the whole process from there, including definitions, code and orchestration. Early this year we decided to build an agentic platform for VaultSpeed, from scratch, and it was the best decision we've taken in years. Today we're announcing VaultSpeed 8, which brings that platform and VaultSpeed's established automation together in one product. It is the moment we bring our founding principle to the market in its agentic form, and we are delivering it in increments, with the capabilities described below arriving release by release.
VaultSpeed 8 is the first release that brings VaultSpeed and the agentic platform together. What follows is what we are building, and each piece reaches you as it is ready.
Modeling with an agent, in a session you can read. You work with an agent on your model and can follow what it proposes and why, step by step, in a session that stays readable afterwards.
An enterprise context store for engineers and agents. Automating the data integration layer produces structured documentation and metadata as a by-product. The context store keeps that information and exposes it over MCP, so an engineer or an agent can ask where a piece of data comes from and how it was integrated, and get a reliable answer.
Data models you maintain elsewhere, converted. You can bring models from other modeling tools, such as Erwin or PowerDesigner, into the context store, so the work you have done carries over.
Code generation that is deterministic where it can be. Where a pattern exists, generation stays deterministic. Where it does not, the work runs as reviewed skills.
Review and versioning in Git. Every agent proposal goes through review and is versioned in Git, so you can see what changed, when, and who accepted it.
We start from the data model because that is where the context lives. An agent that knows how each piece of data was integrated works from a far more reliable base than one that has to infer it.
Deterministic automation keeps handling everything that is already solved, so those problems need no tokens and no LLM sits in the Data Vault code path. Agents take on the specific and unique cases where building a full automation never made sense. They also review each other's work, so a person does not have to approve every single change.
VaultSpeed 8 asks nothing of you. There is no migration and no project to plan, your vaults, metadata and generators keep working as they do today, and you lose nothing of what VaultSpeed already does for you. You can start using the new capabilities as they become available, with nothing to prepare on your side.
A working session with our team is the quickest way to see it. We work directly from your own project and show you what you can do with VaultSpeed 8 on it.
If you are new to VaultSpeed, automating your data integration layer has become much easier to start.
VaultSpeed 8 is the foundation we will keep building on, and the capabilities described above will reach you step by step through the coming releases. Marrying the deterministic engine with agents for the cases that need them is what I'm most excited about today, and I'm looking forward to seeing what you build with it.