mda dev compiles a Managed Deep Agents project and runs it on a local Agent Server. It opens LangSmith Studio so you can interact with the agent and inspect its behavior before deploying.
Managed Deep Agents is in public beta on LangSmith Cloud and Bring Your Own Cloud (BYOC). Use version 0.10 of the Managed Deep Agents CLI and SDK for BYOC.
Start local Studio
Install the project dependencies and add the model provider credentials to.env.
Python projects also require uv.
From the project root, run:
mda dev to recompile the project.
What mda dev does
mda dev:
- Validates the project and compiles it into
.mda/build. - Copies the project
.envinto the local build and adds local-only identity configuration when needed. - Creates a local Context Hub mock for instructions, skills, and memory.
- Starts the language-specific LangGraph development server.
- Opens the agent in Studio.
Trace local runs
mda dev sends traces to the agent’s Local environment. The project name is {name}-local, taken from the name you pass to define_deep_agent or defineDeepAgent.
In an agent-based workspace, LangSmith rejects those traces until the agent record exists. mda deploy creates that record. mda dev alone does not.
Before you rely on local traces, create the agent once with that same identifier in the UI or API, or run mda deploy once. Studio chat still works while traces are rejected.
Configure the local server
For all command details, see the
mda dev CLI reference.
Understand local behavior
mda dev uses local defaults to make testing easier:
- If the agent uses identity, Studio provides a local test user automatically.
- If the configured sandbox is unavailable, the agent uses a temporary local folder instead. The CLI prints the folder path.
- Schedules do not run.
mda devlists each schedule at startup but does not provision it, so no scheduled run fires locally.
Connect these docs to your agent of choice via MCP for real-time answers.

