Understand this story
On November 12, 2026, Mastra launched a private model gateway, allowing Agent requests to be routed to self-hosted open-weight models; simply write the model prefix as private/ollama/gpt-oss:20b to invoke them. According to the Mastra Blog, self-hosting eliminates third parties, prevents sensitive data from leaving the internal network, and costs only infrastructure rather than tokens; setting AUTO_BLOCK_EXTERNAL_PROVIDERS=true in .env makes Studio display only models from your own gateway. This feature requires @mastra/core@1.82.0 or a higher version. There are currently no further developments.
AI-generated from public reports
Latest development · Report
Mastra Launches Private Model Gateway for Agents, Can Route to Self-Hosted Open-Weight ModelsDevelopments
1 confirmed developments
- Source:Mastra Blog
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 12
Mastra BlogFirst-party
Mastra Launches Private Model Gateway for Agents, Can Route to Self-Hosted Open-Weight Models
Mastra has launched a private model gateway, allowing Agent requests to be routed to self-hosted open-weight models; simply write the model prefix as private/ollama/gpt-oss:20b to invoke it. Self-hosting eliminates third parties, prevents sensitive data from leaving the internal network, and costs only infrastructure rather than tokens; setting AUTO_BLOCK_EXTERNAL_PROVIDERS=true in .env makes Studio display only your own gateway models. This feature requires @mastra/core@1.82.0 or higher.