Deep Agent Template (Opinionated, TypeScript)

Opinionated deployment template for a Deep Agent built with createDeepAgent(...).

What this template gives you

• A deployable Deep Agent graph at src/agent.ts.
• Explicit workflow prompt (plan, delegate, critique, finalize).
• Two predefined subagents (researcher, critic).
• Human-in-the-loop interrupts on execute and write_file.
• A Node.js workflow managed by npm with Vitest unit + integration suites.

Quickstart

1. Install dependencies:

npm install

2. Configure environment:

cp .env.example .env

3. Run locally:

npm run dev

Scripts

npm test            # unit tests (src/**/*.test.ts, excludes .int.test.ts)
npm run test:int    # integration tests (requires ANTHROPIC_API_KEY)
npm run test:eval   # evaluation tests with LangSmith reporter
npm run lint        # prettier --check
npm run format      # prettier --write
npm run build       # langgraphjs build

Integration tests are skipped unless ANTHROPIC_API_KEY is set.

Deploy to LangSmith

1. Push this template to a Git repository.
2. In LangSmith, create a new Deployment from that repo.
3. Set environment variables for your selected model provider and optional tracing key.
4. Deploy using the provided langgraph.json.

Reference docs

• Deep Agents (JS): https://github.com/langchain-ai/deepagentsjs
• LangGraph deployment in LangSmith: https://docs.langchain.com/oss/javascript/langchain/deploy