Skip to main content
Glama
duplicati-mcp

duplicati-mcp

by kcofoni

Duplicati MCP Server

MCP (Model Context Protocol) server for managing Duplicati backups from an LLM.

Version française / French version

Architecture

The server wraps the Duplicati REST API and exposes it via the MCP protocol. Two transports are supported:

  • stdio — for local use via Claude Code (no network, no port)

  • Streamable HTTP — for Docker deployment, accessible over the network

Related MCP server: mcp-cli-catalog

Getting Started

Local use with Claude Code (stdio)

The simplest way to get started. The .mcp.json at the project root handles everything:

# Install uv if needed
brew install uv

# Claude Code will auto-detect .mcp.json and launch the server

Set your Duplicati URL and password in .mcp.json:

{
  "mcpServers": {
    "duplicati": {
      "type": "stdio",
      "command": "uv",
      "args": ["run", "duplicati-mcp"],
      "env": {
        "DUPLICATI_URL": "http://localhost:8200",
        "DUPLICATI_PASSWORD": "your-password",
        "DUPLICATI_READONLY": ""
      }
    }
  }
}

With Docker Compose (Docker Hub image)

# Edit DUPLICATI_URL and DUPLICATI_PASSWORD in docker-compose.yml, then:
docker compose up -d

With Docker Compose (local build)

# Edit docker-compose.yml: comment out `image:` and uncomment `build: .`
docker compose up -d --build

Direct Docker usage

docker run -d \
  --name duplicati-mcp-server \
  -p 3000:3000 \
  -e DUPLICATI_URL=http://your-duplicati-host:8200 \
  -e DUPLICATI_PASSWORD=your-password \
  kcofoni/duplicati-mcp:latest

Verification

# Check that the server is running
docker logs duplicati-mcp-server

# Test the MCP endpoint
curl -X POST http://localhost:3000/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-03-26","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}'

Client Configuration

Claude Code — local (stdio)

For local use without Docker, add to your project .mcp.json:

{
  "mcpServers": {
    "duplicati": {
      "type": "stdio",
      "command": "uv",
      "args": ["run", "duplicati-mcp"],
      "env": {
        "DUPLICATI_URL": "http://localhost:8200",
        "DUPLICATI_READONLY": ""
      }
    }
  }
}

Credentials are loaded from the .env file at the project root (see Getting Started).

Claude Code — Docker/remote (HTTP)

Add to your .mcp.json:

{
  "mcpServers": {
    "duplicati": {
      "type": "http",
      "url": "http://your-host:3000/mcp"
    }
  }
}

Claude Desktop

Claude Desktop requires mcp-proxy as a bridge to HTTP servers. Add to your configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "duplicati": {
      "command": "uvx",
      "args": ["mcp-proxy", "--transport", "streamablehttp", "http://your-host:3000/mcp"]
    }
  }
}

Available Tools

Once connected, the LLM has access to:

Backup Jobs

  1. list_backups — List all configured jobs with ID, name, last run date and result

  2. get_backup — Get detailed information about a specific job

  3. run_backup — Trigger a backup job immediately

  4. abort_backup — Abort the currently running backup for a job

Status & Progress

  1. get_progress — Live progress of the active backup task (phase, %, file counts)

  2. get_server_status — Duplicati server state, version and active task

Configuration

  1. export_backup_config — Export a job configuration as JSON

  2. update_backup_config — Update an existing job configuration in place (use with export_backup_config to modify sources, settings, schedule, etc.)

  3. import_backup_config — Import a job configuration from JSON (creates a new job)

History & Diagnostics (SQLite — requires DUPLICATI_DB_PATH)

  1. db_get_backup_metadata — Rich metadata from the local database: last run date, duration, file counts, quota usage, last error

  2. db_get_backup_schedule — Schedule configuration for a backup job

  3. db_list_errors — Recent error log entries, optionally filtered by job

  4. db_list_notifications — System notifications (update alerts, etc.)

  5. db_get_backup_options — Configuration options for a job (compression, retention policy, etc.) — passphrases excluded

  6. db_list_operations — Operation history for a job (Backup, Restore, List, etc.) with timestamps

  7. db_get_operation_log — Full result and statistics for a specific operation

  8. db_list_filesets — Available restore points (backup versions) for a job

Example Prompts

Once the server is connected to your LLM, here are prompts you can use:

General status

  • "What backup jobs are configured on my Duplicati?"

  • "What was the last backup that ran and what was the result?"

  • "Is a backup currently running?"

History & statistics (requires DUPLICATI_DB_PATH)

  • "Show me the last 10 operations for backup job 2"

  • "What is the average duration of recent backups?"

  • "Have there been any errors on my backups in the past few weeks?"

  • "How many files are backed up and what is the total size on the destination?"

Restore points (requires DUPLICATI_DB_PATH)

  • "What restore points are available for my backup job?"

  • "What is the oldest backup available for a restore?"

Configuration (requires DUPLICATI_DB_PATH)

  • "What retention policy is configured on my backup job?"

  • "What compression and encryption options are in use?"

Diagnostics (requires DUPLICATI_DB_PATH)

  • "Are there any pending system notifications on Duplicati?"

  • "Has my Duplicati encountered any errors recently? Which ones?"

  • "Analyse the last backup and tell me if everything went well"

Open-ended (combines multiple tools)

  • "Give me a full health report on my Duplicati backups"

Environment Variables

Variable

Default

Description

DUPLICATI_URL

http://localhost:8200

URL of the Duplicati instance

DUPLICATI_PASSWORD

(empty)

Duplicati web interface password (leave empty if none set)

DUPLICATI_READONLY

(empty)

Set to true, 1 or yes to disable write operations

DUPLICATI_DB_PATH

(empty)

Path to Duplicati-server.sqlite — enables SQLite-backed history tools

MCP_TRANSPORT

stdio

Transport: stdio or streamable-http

MCP_PORT

3000

Port for Streamable HTTP transport

Read-only Mode

DUPLICATI_READONLY=true disables run_backup, abort_backup, update_backup_config and import_backup_config. All read tools remain active. Useful for safely exploring and analysing backup configurations without any risk of modification.

SQLite Access

Setting DUPLICATI_DB_PATH enables the db_* tools, which read directly from the Duplicati SQLite databases. Access is strictly read-only: databases are opened in read-only mode and copied to memory via the SQLite Online Backup API before any query — the live Duplicati databases are never locked or modified.

Local use — point to the server database on your machine:

DUPLICATI_DB_PATH=/path/to/duplicati/config/Duplicati-server.sqlite

Docker — share the Duplicati config directory as a read-only volume. In docker-compose.yml:

services:
  duplicati-mcp:
    # ...
    volumes:
      - duplicati_config:/duplicati-config:ro   # named volume (recommended)
      # or: - /srv/duplicati/config:/duplicati-config:ro  # bind mount
    environment:
      - DUPLICATI_DB_PATH=/duplicati-config/Duplicati-server.sqlite

volumes:
  duplicati_config:   # must be the same volume used by the Duplicati container

Docker Hub

docker pull kcofoni/duplicati-mcp:latest

Development

File Structure

duplicati-mcp/
├── src/
│   └── duplicati_mcp/
│       ├── __init__.py
│       ├── __main__.py
│       ├── client.py        # Duplicati REST API client
│       ├── db.py            # Read-only SQLite access (server DB + per-backup DBs)
│       └── server.py        # FastMCP server and tools
├── mcp-publication/         # MCP registry publication files
├── requirements.txt         # Python dependencies
├── pyproject.toml           # Project metadata
├── Dockerfile
├── docker-compose.yml
├── .mcp.json                # Claude Code local config (stdio)
├── test_server.sh           # Docker container smoke test
├── test_mcp.py              # MCP protocol test
├── README.md                # This file (English)
└── README_fr.md             # French documentation

Running Tests

# Smoke test (requires running Docker container)
./test_server.sh

# MCP protocol test (requires running server)
python test_mcp.py
python test_mcp.py localhost:3000

Interactive Tool Testing (local)

uv run mcp dev src/duplicati_mcp/server.py

Troubleshooting

Cannot connect to Duplicati

Check that DUPLICATI_URL is reachable from the container. If both run in Docker, put them on the same network and use the service name as hostname.

Authentication failed

Verify DUPLICATI_PASSWORD matches the password set in Duplicati's web interface. Leave empty if no password is configured.

MCP endpoint not responding

docker ps | grep duplicati-mcp-server
docker logs duplicati-mcp-server

License

This project is licensed under the MIT License — see the LICENSE file for details.

Available Tools

17 tools
abort_backupB

Abort the currently running backup task for a job.

Args: backup_id: Numeric ID of the backup job to abort.

Inputs: backup_id (required). Returns: result

db_get_backup_metadataA

Return rich metadata for a backup job from the local SQLite database.

Includes last run date, duration, file counts, quota usage, and last error. More detailed than what the REST API exposes.

Args: backup_id: Numeric ID of the backup job (use list_backups to find IDs).

Inputs: backup_id (required). Returns: result

db_get_backup_optionsA

Return the configuration options for a backup job from the local SQLite database.

Includes compression, encryption module, retention policy, file exclusions, etc. Sensitive values such as passphrases are excluded.

Args: backup_id: Numeric ID of the backup job.

Inputs: backup_id (required). Returns: result

db_get_backup_scheduleB

Return the schedule configuration for a backup job from the local SQLite database.

Args: backup_id: Numeric ID of the backup job.

Inputs: backup_id (required). Returns: result

db_get_operation_logA

Return the detailed log and result statistics for a specific operation.

Use db_list_operations to find operation IDs.

Args: backup_id: Numeric ID of the backup job. operation_id: Numeric ID of the operation (from db_list_operations).

Inputs: backup_id (required), operation_id (required). Returns: result

db_list_errorsA

List recent error log entries from the local SQLite database.

Args: backup_id: Numeric ID of the backup job (leave empty for all jobs). limit: Maximum number of entries to return (default 20).

Inputs: backup_id, limit. Returns: result

db_list_filesetsA

List available restore points (backup versions) for a backup job.

Args: backup_id: Numeric ID of the backup job. limit: Maximum number of filesets to return (default 20).

Inputs: backup_id (required), limit. Returns: result

db_list_notificationsA

List recent system notifications from the local SQLite database (update alerts, etc.).

Args: limit: Maximum number of entries to return (default 20).

Inputs: limit. Returns: result

db_list_operationsB

List recent operations (Backup, Restore, List, etc.) for a backup job.

Reads from the per-backup SQLite database for detailed operation history.

Args: backup_id: Numeric ID of the backup job. limit: Maximum number of operations to return (default 20).

Inputs: backup_id (required), limit. Returns: result

export_backup_configA

Export a backup job configuration as a JSON string for backup or migration.

Args: backup_id: Numeric ID of the backup job to export.

Inputs: backup_id (required). Returns: result

get_backupA

Get detailed information about a specific backup job.

Args: backup_id: Numeric ID of the backup job (use list_backups to find IDs).

Inputs: backup_id (required). Returns: result

get_progressA

Get the live progress of the currently running backup task (phase, %, file counts).

Inputs: no named parameters. Returns: result

get_server_statusA

Get Duplicati server status: version, program state, active task, and scheduler info.

Inputs: no named parameters. Returns: result

import_backup_configA

Import a backup job from a JSON configuration string (as produced by export_backup_config).

Args: config_json: JSON string of the backup configuration to import.

Inputs: config_json (required). Returns: result

list_backupsA

List all configured Duplicati backup jobs with ID, name, last run date and result.

Inputs: no named parameters. Returns: result

run_backupB

Trigger a backup job to run immediately.

Args: backup_id: Numeric ID of the backup job to run.

Inputs: backup_id (required). Returns: result

update_backup_configA

Update an existing backup job from a modified JSON configuration (as exported by export_backup_config). Use this to modify sources, settings, schedule, filters, etc. on an existing job.

Args: backup_id: Numeric ID of the backup job to update. config_json: Modified JSON string of the backup configuration.

Inputs: backup_id (required), config_json (required). Returns: result

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 17 tools

    Disambiguation5/5

    Each tool has a clear and distinct purpose. The db_* tools are differentiated by the specific data they retrieve (metadata, options, schedule, logs, etc.), and actions like list_backups, run_backup, abort_backup are unambiguous. No two tools overlap significantly.

    Naming Consistency5/5

    All tools use a consistent verb_noun pattern in snake_case. Database-specific tools are prefixed with 'db_', and other tools follow a similar style (e.g., 'list_backups', 'run_backup', 'get_progress'). No mixing of conventions.

    Tool Count4/5

    17 tools is slightly above the typical 'sweet spot' of 3-15, but each tool addresses a specific need for managing Duplicati backups. The count is not excessive and serves the domain well.

    Completeness3/5

    The tool set covers core operations like listing, running, aborting, and configuring backups, but lacks direct restore and delete functionality. While import/export cover configuration migration, creating a new backup from scratch is only possible via import, not a dedicated 'create' tool.

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      An MCP server that publishes CLI tools on your machine for discoverability by LLMs
      4 npm
      1
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      MCP server that reads Duplicacy backup metrics from a Prometheus exporter and exposes them to LLMs for monitoring backup status, progress, and health.
      6
      391 npm
      3
      GPL 3.0