Regulus
Install See the diff

Compare

Regulus vs LangChain alone

LangChain is an agent-building toolkit. Regulus is a compliance plane targeting Google ADK as primary runtime. An alt-runtime module wraps LangChain4j; new work lands on ADK first.

Pick Regulus when

  • You're targeting Google ADK (Java) and need regulator-grade evidence.
  • Your LangChain stack is already producing audit logs that don't round-trip with your GRC tool.
  • You need multi-regulator composition and a canonical Principal model — LangChain leaves both as your problem.

Pick LangChain alone when

  • You're committed to the LangChain ecosystem (Python / TypeScript) and don't want to migrate the runtime.
  • Your obligations are non-regulated — internal tooling, prototyping, research.
  • You need LangChain's specific abstractions (Chains, Agents, Tools as defined by LangChain) more than you need a compliance plane.

What LangChain is #

LangChain (Python primarily, TypeScript / Java via LangChain4j) is an agent-building toolkit. It provides abstractions for chains, tools, agents, memory, and prompt templates. Its sweet spot is rapid prototyping and the Python ecosystem.

LangChain is not trying to be a compliance plane. It doesn’t ship an audit envelope, a residency check, a model-risk tier model, a canonical Principal abstraction, or a GRC adapter. That’s not a criticism — it’s a different scope.

What Regulus is #

Regulus is the compliance plane. It targets Google ADK as its primary runtime; new plugin development lands on ADK first. A retained regulus-ai-llm module wraps LangChain4j as an alternative runtime — the plugin surface there lags ADK and isn’t the recommended path.

The path if you’re on LangChain today #

Two options:

Option A — migrate the runtime to ADK. If you’re starting a new regulated agent and the LangChain choice was incidental, this is the right move. ADK is Google’s official runtime; Regulus’s full plugin surface lights up; the integration story is one Spring Boot starter.

Option B — keep LangChain, use Regulus alternative-runtime. Wire regulus-ai-llm against your LangChain4j build. The policy + privacy

  • audit plugins work. The model-risk plugin works. Some plugins (those that depend on ADK-specific callbacks like ToolConfirmation) don’t yet. New plugin development typically reaches LangChain4j on a release lag.

The right call depends on how committed you are to the LangChain ecosystem and how regulator-facing your timeline is.

Why ADK is the primary target #

Three reasons that decided this in 2026-05:

  1. ADK is Google’s official Vertex AI Agent Engine runtime. Most regulated buyers in our pipeline are GCP-native; ADK is the default-fit.
  2. ADK’s plugin SPI is documented and stable. Building a compliance plane on a documented extension contract beats building on a moving ecosystem target.
  3. Service extensions wrap Google-shipped impls. RegulusVertexAiSessionService genuinely extends VertexAiSessionService. With LangChain, there’s no equivalent — you wrap your own session abstraction.

Where Regulus doesn’t fit #

If you’re building research tooling, an internal prototype, or something that won’t see a regulator — LangChain is fine. The overhead of a compliance plane only pays back when the obligations exist.

Install Regulus More comparisons