deploy-coding-agent

deploy-coding-agent

An autonomous coding agent deployed with deepagents deploy. Given a task description, it plans, implements, tests, and commits changes inside a LangSmith sandbox with full shell access.

Prerequisites

Variable	Description
ANTHROPIC_API_KEY	Claude model access
LANGSMITH_API_KEY	Required for deploy and the LangSmith sandbox

Copy .env.example to .env and fill in both keys.

Deploy

deepagents deploy

The agent is deployed using the config in deepagents.toml. The [sandbox] section provisions a LangSmith coding sandbox with a Python 3.12 image so the agent can run code safely.

What to try

Once deployed, open the agent in LangSmith and send it tasks like:

• "Add a function that reverses a string and write a test for it"
• "Find all TODO comments in the repo and create a summary"
• "Refactor the main module to use dataclasses"

The agent follows a Plan → Implement → Review → Deliver workflow defined in AGENTS.md.

Structure

deploy-coding-agent/
├── AGENTS.md                  # Agent instructions and workflow
├── deepagents.toml            # Deploy config (model, sandbox)
├── deepagents.assistant-scope.toml  # Assistant-scoped config variant
├── mcp.json                   # MCP server config
└── skills/
    ├── code-review/           # Code review skill with lint helper
    ├── coding-prefs/          # Coding style preferences
    └── planning/              # Task planning skill

Query via SDK

from langgraph_sdk import get_client

client = get_client(url="https://<your-deployment-url>")
thread = await client.threads.create()

async for chunk in client.runs.stream(
    thread["thread_id"], "agent",
    input={"messages": [{"role": "user", "content": "Add a hello_world function and test it"}]},
    stream_mode="messages",
):
    print(chunk.data, end="", flush=True)

Find your deployment URL in LangSmith under Deployments. See the LangGraph SDK docs for more.

Resources

• deepagents deploy docs
• LangSmith sandbox docs