Build a scheduled research agent

This tutorial builds a research assistant one capability at a time. Complete the quickstart first to scaffold a project, add API keys, and run mda dev locally. Then add a search tool, use durable memory, run the agent on a daily schedule, and deploy it to LangSmith.

Managed Deep Agents is in public beta and available on LangSmith Cloud in the US region only.

Build the agent

Write the instructions

Replace instructions.md with the research assistant’s behavior. The instructions reference the tool you add next and the shared durable memory you explicitly enable later in this tutorial:

instructions.md

# Research assistant

You are a careful research assistant. Find sources, keep notes, and return
concise answers with citations.

## Behavior

- Use the `web_search` tool to find sources instead of guessing.
- Cite the sources you used.

## Memory

- Record reusable research procedures and project knowledge that can improve future work.
- For release research, check the project's official changelog before secondary sources.
- Never store personal data or secrets in memory.

Add a search tool

Create a tools/ module with a search tool, then import it into the agent entry. This example returns a placeholder result, so it runs without an external API. Replace the body with a call to your search provider.

tools/search.py

from langchain.tools import tool


@tool(parse_docstring=True)
def web_search(query: str) -> str:
    """Search the web for a query and return result snippets.

    Args:
        query: The search query.
    """
    # Replace this stub with a call to your search provider.
    return f"Top results for '{query}': ..."

Import the tool into the agent entry and pass it to the definition:

agent.py

from managed_deepagents import define_deep_agent

from tools.search import web_search

agent = define_deep_agent(
    name="research-assistant",
    model="openai:gpt-5.5",
    tools=[web_search],
)

For more on authored tools, see Custom tools.

Run the agent locally

Install dependencies and start the local dev server:

uv sync
mda dev .

mda dev opens the agent in LangSmith Studio. Send a question and confirm the agent calls web_search and answers with the returned snippets.

Enable and use durable memory

Durable memory is opt-in. Before asking the agent to remember anything, add a memory declaration at the project root:

memory.py

from managed_deepagents import define_memory

memory = define_memory(scope="agent")

Memory is shared across the deployment and visible to all callers, so do not store personal data or secrets.Restart mda dev so it discovers the new file. In one thread, ask the agent to research a release and to record a reusable project rule, such as “For release research, check the official changelog before secondary sources.” Then create a new thread in Studio and ask how it will research the next release. Confirm that it applies the shared rule even though the new thread has no conversation history.See Memory for details.

Schedule a daily digest

Add a schedules/ module so the agent runs on a cron cadence without a user message. This schedule runs every weekday at 8am Pacific:

schedules/daily_digest.py

from managed_deepagents import define_schedule

schedule = define_schedule(
    cron="0 8 * * 1-5",
    timezone="America/Los_Angeles",
    prompt="Summarize what you learned yesterday and list open questions.",
)

mda deploy reconciles this schedule into a LangSmith cron job after the deployment is live. For thread behavior and constraints, see Schedules.

Deploy the agent

Deploy the project to LangSmith:

mda deploy .

On success, the CLI prints the deployment dashboard URL. The deploy syncs the instructions to Context Hub, uploads the compiled project, and reconciles the daily schedule. For deploy flags and troubleshooting, see Deploy an agent and the CLI reference.

Inspect the run

Open the printed URL in LangSmith to inspect build status and revisions. Open traces to inspect the agent’s inputs, model calls, web_search calls, memory reads and writes, and final responses.

Next steps

Custom middleware

Add logging, retries, limits, and guardrails around model and tool calls.

Identity

Authenticate callers and use verified identity in tools and middleware.

Memory

Persist shared procedural and project knowledge across threads.

Evals

Author Harbor tasks and compile the managed agent for Harbor.

Sandboxes

Configure isolated filesystem and shell access for agent work.

Connect these docs to Claude, VSCode, and more via MCP for real-time answers.

Edit this page on GitHub or file an issue.

Get started

Agent definition

Build and deploy

Build a scheduled research agent

Build the agent

Next steps

Custom middleware

Identity

Memory

Evals

Sandboxes

​Build the agent

​Next steps

Custom middleware

Identity

Memory

Evals

Sandboxes

Build the agent

Next steps