MeshAgent 0.40 makes four agent-building workflows easier: starting realtime text and voice conversations, configuring managed agents and routes, scaffolding new projects with meshagent create, and using the same agent and route concepts across Dart, TypeScript, Python, and .NET SDKs.
MeshAgent agents can now carry more of the choices that matter for realtime conversations: provider, model, voice, output modality, audio formats, and realtime protocol. That means a client can start a thread or turn with the model and modality it needs, instead of treating voice and chat behavior as separate from the normal agent protocol.
This matters for apps that mix text chat, voice, and live collaboration inside a room. A Powerboards-style client, a custom Flutter app, or a Python process can ask for text or audio output, use realtime-capable adapters, and keep the resulting turn inside the same thread model that other room participants already use.
The release also improves OpenAI realtime handling, adds Dart audio and WAV protocol support, updates TypeScript websocket behavior for Node, and adds CLI audio dependencies so local process workflows can participate in realtime sessions more directly.
MeshAgent 0.40 expands the managed-agent model. Managed-agent specs now cover allowed models, toolkits, secrets, MCP servers, thread isolation, room and agent grants, and agent session listing.
Route definitions also move toward a shared RouteSpec shape that can describe room and agent backends.
The chat and channel transport gained websocket support, participant connect and disconnect events, sender-name propagation, and better thread start/load behavior when attachments are involved. Those changes make room chat and managed-agent conversations easier to represent in product surfaces that need to show who joined, who sent a message, what files were attached, and which session is active.
meshagent create replaces older starter entrypointsMeshAgent 0.40 introduces a new meshagent create workflow for starting projects and examples.
meshagent create
meshagent create --help
The scaffolder includes templates for Dart, .NET, JavaScript, Python, React, and TypeScript. This makes it easy to start from a template that matches the language or app shape you want to build and quickly see how to run apps with MeshAgent.
Dart adds the new meshagent-agents-dart package, native websocket handling, route-spec APIs, managed-agent secret support, agent route queries, agent session support, and Flutter desktop updater improvements. TypeScript adds route-spec CRUD and listing for room and agent routes, websocket compression support in Node, and realtime protocol fields. Python updates the core agent message and adapter contracts. .NET adds route-spec types and route CRUD/listing APIs for room and agent routes.
The Python agent adapter method formerly named next(...) is now create_response(...), matching the language used by realtime and response-oriented adapters. Toolkit serialization also drops older thumbnailUrl and pricing fields from managed-agent toolkit descriptions.
The ask and process flows also get more collaborative. Thread loading, new-thread creation, multi-user terminal UI behavior, status reporting, and tool-call summaries were reworked so local agent sessions can behave more like participants in a room.
That makes a difference when a developer is debugging an agent with other people in the same room, attaching files, switching threads, or watching tool use unfold. The terminal can show more of the session state that product clients and room participants care about.
Taken together, MeshAgent 0.40 makes richer agent workflows easier to build with realtime audio and text modalities, managed-agent policy, route-backed services, improved starter templates, and SDKs that expose those pieces more consistently.
Join our Discord community to keep up with MeshAgent releases, ask questions, and share feedback with the 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.
