Managed Deep Agents project structure

​

Project layout

my-agent/
├── agent.ts | agent.tsx            # Core agent definition

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

├── tools/                          # Application code
│   └── mcp.ts                      # MCP server declaration
├── middleware/

├── channels/                       # Managed configuration
│   └── <name>.ts
├── schedules/
│   └── <name>.ts
├── sandbox/
│   └── index.ts
├── identity.ts
├── memory.ts

├── package.json                    # Dependencies
├── .env                            # Local and deploy secrets

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

​

How MDA treats project files

• Managed context: instructions.md defines the system prompt. Each directory under skills/ contains task-specific instructions, such as a SKILL.md and any supporting files. MDA syncs both instructions.md and skills/ to Context Hub. Optional durable memory is also backed by Context Hub.
• Application code: Files under tools/ and middleware/ are ordinary project modules. Import them from the agent entry.
• Managed configuration: Certain paths enable capabilities when present. For channels/ and schedules/, only direct children are managed declarations; nested modules are not.

  Path	Enables
  identity.ts	Caller authentication
  memory.ts	Durable memory
  channels/<name>.ts	Messaging channels
  tools/mcp.ts	MCP connectors
  schedules/<name>.ts	Cron schedules
  sandbox/index.ts	Sandbox filesystem and shell

  tools/ holds ordinary modules with one exception. tools/mcp.ts is a managed declaration and exports a named mcp. Every other module under tools/ is application code you import.
• Dependencies and secrets: Declare dependencies in package.json. MDA loads .env locally and forwards non-reserved values as deployment secrets. Reserved platform variables and .env files are not included in the build archive. For more information, see Deploy a Managed Deep Agent.
• 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.

​

Next steps