MeshAgent 0.36 adds three practical workflow upgrades: service specs can describe more of how an agent runs, queue jobs can send richer input into agents, and Studio now gives teams direct room registry management workflows.
This release builds on existing deployment and runtime workflows. Across the SDKs, service definitions can now include typed prompt content, agent email and heartbeat settings, service files, and richer container mounts. Queue-driven work can now carry structured content and room-file references, and Studio plus Dart clients now make it easier to inspect and manage room registry container images after they have been built and published.
MeshAgent 0.36 adds several practical pieces to service definitions.
Service definitions can now carry typed prompt content as explicit text and file items instead of treating prompt input as only a raw string. Service models also pick up agent email and heartbeat settings, service files, and richer container storage options including config mounts and empty directories. Service-template conversions and editors also now include config and empty-dir mounts alongside the existing project, image, and file mount options.
That matters when teams want to describe more of a service's runtime shape directly: what prompt content it starts with, what files or runtime metadata it needs, and how it should participate in email or scheduled heartbeat workflows.
This release also removes container API key provisioning from container specs. The overall direction is a clearer boundary between container shape and the credentials or runtime configuration that surround it.
MeshAgent 0.36 also improves what a queue-driven workflow can send into an agent.
Queue channels now accept structured `content` and `prompt` payloads, including `room://` file references. Legacy prompt-file flows still resolve, but queue messages no longer need to collapse everything into one plain-text field. Queue-driven workflows also gain thread ID templates with time tokens, which makes it easier to route repeated jobs into predictable thread patterns.
In practice, this is useful when queue work is more than a one-line instruction. A scheduled job can point at room files directly, pass multiple typed content items, and keep repeated runs grouped into predictable thread IDs instead of creating disconnected context every time.
Shell-based workflows also get a useful runtime bridge in this release. With `--shell-tool-config-mount`, a shell container can receive MeshAgent runtime metadata such as `spec.json` and `members.json` as mounted files. That gives agents and operator workflows a direct way to inspect room or service runtime state from inside a managed shell environment.
Teams building and deploying room services also get new room registry management APIs and UI in 0.36.
The room registry APIs now support listing repositories, inspecting tags, retagging images, and deleting room-registry refs from Dart and Flutter clients, and MeshAgent Studio's Registry tab adds UI for those same workflows. That means teams can inspect what has already been published into a room, promote a tag, or clean up old refs without leaving the MeshAgent environment.
This is especially useful for rooms that build, test, and promote service images frequently. The registry becomes something teams can work with directly from MeshAgent and does not require you to push to external container registries.
The Flutter and Studio layers also pick up smaller but useful quality-of-life improvements in this release. Attachment-only chat messages stay visible, unsupported event kinds are filtered more consistently, and the service-template editor now defaults enum fields to valid values and normalizes invalid selections automatically.
Join our Discord Server to stay up to date with MeshAgent releases, see full release notes, ask questions, and share feedback with our team.
Check out the MeshAgent documentation to start building today!
MeshAgent Studio, SDK, and Server give you everything to build, test, and deploy agentic applications, from development to production.
