🗄️ Deep Agents Remote Backends

deepagents-backends provides production-ready implementations of the LangChain Deep Agents' BackendProtocol for remote file storage, allowing your agents to maintain state across restarts and share files in distributed environments.

Store agent files in S3 or PostgreSQL instead of ephemeral state, enabling persistent storage, distributed execution, and multi-agent file sharing.

🚀 Quickstart

pip install deepagents-backends

S3 Backend

Store agent files in AWS S3 or any S3-compatible storage (MinIO, DigitalOcean Spaces, etc.):

import asyncio

from deepagents import create_deep_agent
from deepagents_backends import S3Backend, S3Config
from langchain_anthropic import ChatAnthropic


def create_default_model() -> ChatAnthropic:
    return ChatAnthropic(
        model_name="claude-sonnet-4-5-20250929",
        max_tokens=20000,
        betas=["prompt-caching-2024-07-31"],
    )


async def main():
    config = S3Config(
        bucket="my-agent-bucket",
        prefix="agent-workspace",
        endpoint_url="http://localhost:9000",  # Remove for AWS S3
        access_key_id="minioadmin",
        secret_access_key="minioadmin",
        use_ssl=False,
    )

    agent = create_deep_agent(
        model=create_default_model(),
        backend=S3Backend(config),
        system_prompt="You are a helpful assistant. Files persist in S3.",
    )

    result = await agent.ainvoke({
        "messages": [{"role": "user", "content": "Create a Python calculator module in /src/"}]
    })

    print(result)


asyncio.run(main())

PostgreSQL Backend

Store agent files in PostgreSQL with connection pooling for high-performance scenarios:

import asyncio
import sys

from deepagents import create_deep_agent
from deepagents_backends import PostgresBackend, PostgresConfig
from langchain_anthropic import ChatAnthropic

if sys.platform == "win32":
    asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())


def create_default_model() -> ChatAnthropic:
    return ChatAnthropic(
        model_name="claude-sonnet-4-5-20250929",
        max_tokens=20000,
        betas=["prompt-caching-2024-07-31"],
    )


async def main():
    config = PostgresConfig(
        host="localhost",
        port=5432,
        database="deepagents",
        user="postgres",
        password="postgres",
        table="agent_files",
    )

    backend = PostgresBackend(config)
    await backend.initialize()

    try:
        agent = create_deep_agent(
            model=create_default_model(),
            backend=backend,
            system_prompt="You are a data analyst. Files persist in PostgreSQL.",
        )

        result = await agent.ainvoke({
            "messages": [{"role": "user", "content": "Create a data analysis project in /analysis/"}]
        })
    finally:
        await backend.close()

asyncio.run(main())

🔀 Composite Backend (Hybrid Storage)

Route different paths to different backends for optimal storage:

import asyncio
import sys

from deepagents import create_deep_agent
from deepagents.backends import CompositeBackend, StateBackend
from deepagents_backends import PostgresBackend, PostgresConfig, S3Backend, S3Config
from langchain_anthropic import ChatAnthropic

if sys.platform == "win32":
    asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())


def create_default_model() -> ChatAnthropic:
    return ChatAnthropic(
        model_name="claude-sonnet-4-5-20250929",
        max_tokens=20000,
        betas=["prompt-caching-2024-07-31"],
    )


async def main():
    s3_backend = S3Backend(
        S3Config(
            bucket="my-asset-bucket",
            prefix="agent-assets",
            region="us-east-1",
        )
    )
    pg_backend = PostgresBackend(
        PostgresConfig(
            host="localhost",
            port=5432,
            database="deepagents",
            user="postgres",
            password="postgres",
            table="agent_files",
        )
    )
    await pg_backend.initialize()

    try:
        agent = create_deep_agent(
            model=create_default_model(),
            backend=lambda runtime: CompositeBackend(
                default=StateBackend(runtime),
                routes={
                    "/assets/": s3_backend,
                    "/data/": pg_backend,
                    "/memories/": pg_backend,
                },
            ),
        )

        await agent.ainvoke({
            "messages": [{"role": "user", "content": "Set up a hybrid workspace under /assets and /data."}]
        })
    finally:
        await pg_backend.close()


asyncio.run(main())

📚 Examples

See the examples/ directory for complete, runnable examples:

Example	Description
s3_deep_agent.py	Full S3 backend with streaming and custom tools
postgres_deep_agent.py	PostgreSQL with multi-agent and sub-agent workflows
composite_backend.py	Hybrid S3 + PostgreSQL storage with routing
basic_usage.py	Low-level backend API operations

Running Examples Locally

# Start MinIO and PostgreSQL
docker-compose up -d

# Run an example
uv run examples/s3_deep_agent.py

⚙️ Configuration

S3Config

S3Config(
    bucket="my-bucket",              # Required: S3 bucket name
    prefix="agent-files",            # Key prefix for all files
    region="us-west-2",              # AWS region (default: us-east-1)
    endpoint_url=None,               # Custom endpoint (MinIO, etc.)
    access_key_id=None,              # AWS credentials (or use IAM role)
    secret_access_key=None,
    use_ssl=True,                    # Use HTTPS
    max_pool_connections=50,         # Connection pool size
    connect_timeout=5.0,             # Connection timeout (seconds)
    read_timeout=30.0,               # Read timeout (seconds)
    max_retries=3,                   # Retry attempts
)

PostgresConfig

PostgresConfig(
    host="localhost",                # PostgreSQL host
    port=5432,                       # PostgreSQL port
    database="deepagents",           # Database name
    user="postgres",                 # Username
    password="postgres",             # Password
    table="agent_files",             # Table name for file storage
    min_pool_size=5,                 # Minimum connections in pool
    max_pool_size=20,                # Maximum connections in pool
    sslmode="prefer",                # SSL mode (use "require" in production)
)

🔧 Backend Protocol

Both backends implement the full BackendProtocol with sync and async methods:

Method	Description
read / aread	Read file content (supports offset/limit pagination)
write / awrite	Create new file (fails if exists)
edit / aedit	Edit file with string replacement
ls_info / als_info	List directory contents
glob_info / aglob_info	Find files matching glob pattern
grep_raw / agrep_raw	Search files with line-numbered results
upload_files / aupload_files	Batch upload raw bytes
download_files / adownload_files	Batch download as bytes

File Storage Format

Files are stored as JSON with line arrays for efficient line-based operations:

{
  "content": ["line 1", "line 2", "line 3"],
  "created_at": "2025-01-07T12:00:00Z",
  "modified_at": "2025-01-07T12:30:00Z"
}

🧪 Development

# Install dev dependencies
uv sync

# Unit tests (mocked, no Docker)
uv run pytest -m unit

# Integration tests (Docker services started automatically via pytest-docker)
uv run pytest -m integration

# All tests
uv run pytest

Docker Services

Service	Port	Credentials
MinIO (S3)	9000	minioadmin / minioadmin
MinIO Console	9001	minioadmin / minioadmin
PostgreSQL	5432	postgres / postgres

🔒 Security

• Credentials: Use environment variables or IAM roles, never commit secrets
• PostgreSQL: Use sslmode="require" in production
• S3: Use use_ssl=True in production
• Connection pooling: PostgresBackend maintains a connection pool—always call close()

📄 License

MIT License - see LICENSE for details.