Introduction

Sparq is the governed layer between your applications and AI models. Your application never calls a model directly. It talks to a Sparq Project over one versioned contract, and Sparq does the rest: it runs the agents, routes every model call to a deployment your policy allows, calls your tools, asks a person when an action needs approval, and records who did what, with which credential, at what cost.

Your application ──▶ Sparq Project ──▶ Frontier Gateway ──▶ models
   (REST + SSE)        (agents, policy,     (by model class,
                        budgets, audit)      filtered by policy)
                            │
                            └──▶ tool gateway ──▶ your tools (MCP, your APIs)

Because the contract is with the Project, the models, instructions, tools and knowledge behind it can change, upgrade or roll back without your application changing a line of code.

How these docs are organisedCopy Anchor Link

StatusCopy Anchor Link