Mastra Factory ships new capabilities: an open-source software factory that starts real work within five minutes, as weekly merged PRs per engineer rise from 9.5 to 16.9
Synopsis
Mastra released a set of new capabilities for its open-source software factory Mastra Factory, including multiple intake sources, customizable workflow stages, a cloud coding agent, integrated workspaces and flexible connectors, with support for any model provider; the company says a factory can be set up, customized and deployed to start doing real work in under five minutes. Factory had been in alpha and beta, and internal trust was uneven: through early October no more than three engineers ran Factory sessions in a given week. Mastra's own usage figures put merged PRs per engineer per week at 16.9 over the last four complete weeks, up 78% from 9.5 before Factory, though the release is not yet a 1.0.0 version.
Interpretation
Mastra Factory now offers a software factory teams can set up, customize and deploy, which the company says lets them start doing real work in under five minutes, with work moving from intake to shipped through customizable workflow stages. In the alpha and beta phases Factory was usable, but trust inside the team was inconsistent and a handful of power users accounted for most of the usage. Mastra's own figures are 9.5 merged PRs per engineer per week before Factory and 16.9 over the last four complete weeks, a 78% increase.
Factory intake comes from sources such as GitHub, GitLab, Linear, Jira or incidents from systems like Incident.io, and work items advance through stages such as Triage, Plan, Review and Done, each bindable to a team's own skills, tooling and human-approval rules. Coding is handled by a cloud coding agent backed by Mastra Code, carrying observational memory persisted across work items on the board. Every session gets a microVM, with E2B the default on the Mastra platform, so work runs as close as possible to the team's development environment.
Model access is not tied to one provider: any model provider and gateway, inference provider, frontier lab or OpenAI-compatible model shape is supported, and personal, hybrid and org modes decide how models are used. Teams can bring their own gateway, inference provider or preferred model family, and can attach custom model gateways where data privacy and compliance restrictions apply. Mastra also reports work-item figures: in August a new work item sat in triage for a median of 6.9 hours, and in November it is 11 minutes, with the Factory agent doing 91% of the triage.
Perspective
The demonstrated scope is the Mastra team's use on its own codebase and its own factory, Shipyard, together with the intake sources, workflow stages, cloud coding agent, integrated workspaces, connectors and model access this release provides; the intended reader is an engineering team looking to bring coding agents into its delivery process.
Mastra states that this is not a 1.0.0 release and that Factory will keep evolving and hardening; the reported merged-PR counts, triage times and merge times are the team's own usage statistics, and the company also notes that engineering remains a tool-assisted craft and expects to use factories further.
