Managed Deep Agents project structure

​

Project layout

my-agent/
├── agent.py                        # Core agent definition

├── instructions.md                 # Managed context
├── skills/
│   └── <name>/
│       └── SKILL.md

├── tools/                          # Application code
├── middleware/

├── channels/                       # Managed configuration
│   └── <name>.py
├── connectors/
│   └── mcp.py
├── schedules/
│   └── <name>.py
├── sandbox/
│   └── __init__.py
├── identity.py
├── memory.py

├── pyproject.toml                  # Dependencies and secrets
├── .env

└── evals/                          # Harbor workspace
    ├── harbor-job.json
    └── <task>/                     # Harbor task
        ├── Task.md
        ├── instruction.md
        ├── environment/
        └── tests/

​

How Managed Deep Agents treats project files

• Managed context: instructions.md defines the system prompt. Each directory under skills/ contains task-specific instructions. Managed Deep Agents syncs both to Context Hub.
• Application code: Files under tools/ and middleware/ are ordinary project modules. Import them from the agent entry. Other local modules work the same way.
• Managed configuration: Root identity.py and memory.py, direct children of channels/, connectors/, and schedules/, and sandbox/__init__.py enable their corresponding capabilities. MCP connector modules export a module-level connector.
• Dependencies and secrets: Declare dependencies in pyproject.toml. Managed Deep Agents loads .env locally and forwards eligible values as deployment secrets, but never includes .env files in the build archive.
• Evals: Managed Deep Agents evals are Harbor evals. Run mda evals init -i and develop tasks with a coding agent and the eval-engineering skill. Generated runtime files stay under .mda/evals/ and are not included in the deployed agent build.