gram
View on GitHubSecurely scale AI usage across your organization. A single stack to Connect, Secure, Observe and Distribute agents, MCPs, and Skills within your company.
Gram is Speakeasy's open source AI control plane: a self-hostable gateway to connect, secure, observe and distribute MCP servers, Skills and assistants across an organization, with RBAC, SSO sync, PII/injection guardrails, token-cost telemetry and full audit logging.
Use Cases
Centrally distribute MCP servers, Skills and assistants to employeesEnforce RBAC and OAuth2.1 permissions on agent toolsTrack AI token usage and cost per team or sessionBlock or redact PII and secrets in prompts and responsesDetect prompt injection and shadow AI usageConnect agents to SaaS, internal APIs and systems through one control layerSync access with Okta, Azure AD and Google WorkspaceAudit and search every tool call and permission changeRun private networking and tunneling for agent trafficSet budgets and analyze expensive agent sessions
Built With
- Language
- Go
- Frameworks
- Goa · Temporal · OpenRouter · OpenTelemetry · React · TypeScript · Vite · ClickHouse · Redis · Docker · Mise · Playwright · Polar · Go
Tags
mcp-gateway · mcp-server · mcp-tools · ai-control-plane · agents · skills · observability · rbac · security · llm-gateway · openapi · opentelemetry · guardrails · sso · token-cost-tracking · governance