arcticdb-mcp 0.1.3

MCP server that gives AI assistants full read/write access to ArcticDB

arcticdb-mcp

MCP server that gives AI assistants full read/write access to ArcticDB — versioned, queryable, serverless DataFrames at scale.

---

Why ArcticDB + AI

Most databases give your AI assistant data. ArcticDB gives it versioned data — every write is a new version, every version is recoverable. This means your AI can:

• Detect when a new data update introduced anomalies by comparing it to the previous version
• Roll back to a known-good state without any data loss
• Query billions of rows with date range, filter, and groupby — without loading them into memory
• Work across LMDB (local), S3, Azure Blob, and MinIO with zero infrastructure changes

---

Configure Your AI Assistant

We provide full instructions for configuring arcticdb-mcp with Claude Desktop. Many MCP clients use a similar configuration file — you can adapt these steps to work with the client of your choice.

Cursor / Windsurf / Continue.dev — these all use the same JSON config format as Claude Desktop. Use any of the blocks below as-is.

n8n / HTTP-based tools — see HTTP / SSE mode below.

---

Claude Desktop configuration file location

OS	Path
macOS	~/Library/Application Support/Claude/claude_desktop_config.json
Windows	%APPDATA%\Claude\claude_desktop_config.json
Linux	~/.config/Claude/claude_desktop_config.json

You can also use the Settings menu in Claude Desktop to locate the file.

Edit the mcpServers section using one of the methods below.

---

If you are using uvx (recommended)

No install step needed — uvx fetches and runs arcticdb-mcp directly from PyPI.

{
  "mcpServers": {
    "arcticdb": {
      "command": "uvx",
      "args": ["arcticdb-mcp"],
      "env": {
        "ARCTICDB_URI": "lmdb:///path/to/your/database"
      }
    }
  }
}

---

If you are using uv

{
  "mcpServers": {
    "arcticdb": {
      "command": "uv",
      "args": ["run", "arcticdb-mcp"],
      "env": {
        "ARCTICDB_URI": "lmdb:///path/to/your/database"
      }
    }
  }
}

---

If you are using pipx

pipx install arcticdb-mcp

{
  "mcpServers": {
    "arcticdb": {
      "command": "arcticdb-mcp",
      "env": {
        "ARCTICDB_URI": "lmdb:///path/to/your/database"
      }
    }
  }
}

---

If you are using pip

pip install arcticdb-mcp

{
  "mcpServers": {
    "arcticdb": {
      "command": "python",
      "args": ["-m", "arcticdb_mcp"],
      "env": {
        "ARCTICDB_URI": "lmdb:///path/to/your/database"
      }
    }
  }
}

---

If you are using Docker

Local LMDB — mount your database directory as a volume:

{
  "mcpServers": {
    "arcticdb": {
      "command": "docker",
      "args": [
        "run", "--rm", "-i",
        "-e", "ARCTICDB_URI=lmdb:///data/db",
        "-v", "/path/to/your/database:/data/db",
        "ymuskrat/arcticdb-mcp"
      ]
    }
  }
}

S3 / Azure — no volume needed, pass credentials via env:

{
  "mcpServers": {
    "arcticdb": {
      "command": "docker",
      "args": ["run", "--rm", "-i", "ymuskrat/arcticdb-mcp"],
      "env": {
        "ARCTICDB_URI": "s3://s3.amazonaws.com:bucket?region=us-east-1&access=KEY&secret=SECRET"
      }
    }
  }
}

Build the image locally:

docker build -t arcticdb-mcp .

---

ArcticDB URI

Replace lmdb:///path/to/your/database with your ArcticDB connection URI:

Backend	URI format
Local (LMDB)	lmdb:///path/to/db (Linux/Mac) · lmdb://C:/path/to/db (Windows)
AWS S3	s3://s3.amazonaws.com:bucket?region=us-east-1&access=KEY&secret=SECRET
Azure Blob	azure://AccountName=X;AccountKey=Y;Container=Z
MinIO / S3-compatible	s3://your-endpoint:bucket?access=KEY&secret=SECRET

You can also set the URI in a .env file in your working directory instead of the config.

---

Then ask your AI assistant:

"Show me the last 5 rows of AAPL in the finance library"
"Filter NVDA prices greater than 500 in 2024"
"Compare the latest version of this symbol to the previous one"
"Create a snapshot of the finance library before I update it"

---

Backends

Backend	URI format
Local (LMDB)	lmdb:///path/to/db (Linux/Mac) · lmdb://C:/path/to/db (Windows)
AWS S3	s3://s3.amazonaws.com:bucket?region=us-east-1&access=KEY&secret=SECRET
Azure Blob	azure://AccountName=X;AccountKey=Y;Container=Z
MinIO / S3-compatible	s3://your-endpoint:bucket?access=KEY&secret=SECRET

Set the URI via environment variable:

export ARCTICDB_URI="lmdb:///path/to/your/database"

Or use a .env file in your working directory.

---

Tools

Libraries

Tool	Description
list_libraries	List all libraries in the Arctic instance
create_library	Create a new library
delete_library	Delete a library and all its data
library_exists	Check whether a library exists
get_library	Get a library's name and symbol list

Symbols

Tool	Description
list_symbols	List all symbols in a library
read_symbol	Read a symbol's full data. Use as_of for a specific version
head_symbol	Read the first n rows of a symbol
tail_symbol	Read the last n rows of a symbol
write_symbol	Write data to a symbol, creating a new version
append_symbol	Append rows to an existing symbol
update_symbol	Overwrite a date range of a timeseries symbol
delete_symbol	Delete a symbol and all its versions
delete_data_in_range	Delete rows within a date range, creating a new version
symbol_exists	Check whether a symbol exists
get_symbol_info	Get row count, column names, and last update time
read_metadata	Read a symbol's metadata without loading data
write_metadata	Update a symbol's metadata
list_versions	List all versions of a symbol with timestamps

Snapshots

Tool	Description
create_snapshot	Snapshot the current state of a library
list_snapshots	List all snapshots in a library
delete_snapshot	Delete a named snapshot
read_symbol_from_snapshot	Read a symbol as it existed at snapshot time

Queries

Tool	Description
query_filter	Filter rows with >, >=, <, <=, ==, != conditions
query_filter_isin	Filter rows where a column value is in a list
query_groupby	Group by a column and aggregate (mean, sum, min, max, count)
query_date_range	Filter a datetime-indexed symbol by date range
query_resample	Resample a datetime-indexed symbol (1h, 1D, 1W, ...)

---

HTTP / SSE mode (n8n and other tools)

Tools like n8n, Zapier, or any custom application that can't launch a local process need the server running as an HTTP service. Set ARCTICDB_MCP_PORT to switch from stdio to SSE transport:

ARCTICDB_URI=lmdb:///path/to/your/database ARCTICDB_MCP_PORT=8000 python -m arcticdb_mcp

The MCP endpoint will be available at:

http://localhost:8000/sse

With Docker:

docker run -d \
  -e ARCTICDB_URI=lmdb:///data/db \
  -e ARCTICDB_MCP_PORT=8000 \
  -v /path/to/your/database:/data/db \
  -p 8000:8000 \
  ymuskrat/arcticdb-mcp

In n8n — add an MCP Client node and point it to http://your-host:8000/sse.

---

Development

git clone https://github.com/YMuskrat/arcticdb-mcp
pip install -e ".[dev]"

# Test with MCP Inspector
ARCTICDB_URI=lmdb:///tmp/test_db npx @modelcontextprotocol/inspector python -m arcticdb_mcp

---

License

MIT