# Gram

An open-source control plane for centrally managing MCPs, agent skills and assistants across a company, adding fine-grained permissions, real-time policy enforcement over prompts and responses, and full observability of tool calls, token use and cost.

Gram is the open-source stack behind Speakeasy's AI control plane, meant to let an organization securely scale AI usage rather than have each team wire up its own MCP connections. Observability tracks AI usage by team, in tokens or cost, lets an operator drill into an expensive session, set budgets, and measure how effective a given tool is, built on OpenTelemetry and exportable through the platform's MCP server or a built-in assistant. A security layer inspects every prompt, response and agent action in real time, blocking, redacting or logging sensitive data before it leaves the environment under configurable policies aimed at prompt injection, shadow use and PII exposure.

A connect layer gives agents a single control point for reaching MCPs across SaaS vendors, internal APIs and other systems, with policy enforcement, granular access control and on-demand private networking or tunneling. A distribute layer centralizes rollout of MCPs, skills, plugins and assistants to a team, enforced through role-based access control and OAuth 2.1, synced to an identity provider such as Okta, Azure AD or Google Workspace. Running it locally uses a single setup script that installs dependencies, applies database migrations, and starts the dashboard and services, built on Go, TypeScript and Mise plus Docker.

Website: https://www.speakeasy.com