Govern Gemini agents before the first step is emitted

Compile SOPs into Gemini API configurations, validate against ETHOS.md, and simulate cost before the first token is spent.

Try the Demo

Google AI Studio (Gemini Interactions) Governance | Cromus

Google announced the Gemini Interactions API (May 2026) — instead of strict "user" and "model" roles, every action is now represented as its own typed step. The seven step types are user_input, thought, function_call, function_call_result, google_search_call, google_search_result, and model_output.

Cromus is the pre-execution governance layer for the Gemini Interactions API. Cromus compiles SOPs into SKILL.md capability specs, validates them against ETHOS.md behavioral policies, simulates cost across the Gemini model family (gemini-2.5-pro, gemini-2.5-flash, gemini-2.5-flash-lite, gemini-2.0-pro, gemini-2.0-flash), and produces a Google AI Studio configuration object — model, system_instruction, tools — plus a typed steps-response template that maps to the new Interactions API contract.

The four pre-execution steps are: Compile (SOP into SKILL.md), Validate (SKILL.md against ETHOS.md), Simulate (cost and Croms score across Gemini pricing), and Deploy (compile to JSON config for Google's Gemini API). Each step produces an audit-ready artifact that travels with the agent. Cromus normalizes and whitelists Gemini-supported tools (google_search, code_execution, function_declarations, url_context) and emits the steps-response template so Interactions API consumers know exactly which step types to handle.

Cromus's three open specs do specific jobs in this flow. SKILL.md is the capability definition. ETHOS.md is the behavioral policy. MEMORY.md is the portability contract. All three are open, free, and validator-checkable. They are designed to ingest into any agent runtime — Google AI Studio, Claude Managed Agents, Replit Agents, AgentSkills.io, OpenClaw, custom infrastructure. The point of an open spec is that it survives the platform.

Runtime governance secures agents that are already deployed. Pre-execution governance happens before any of that — before the SOP becomes a skill, before the skill becomes an agent definition, before the agent definition becomes a deployed Gemini config. Cromus sits upstream of any runtime, including Google AI Studio.