Firebird

You can use the Firebird MCP Server to inspect Firebird database schemas and run read-only queries through a lightweight MCP client. It exposes table schemas and a query tool so you can safely explore data structures without modifying anything.

How to use

You interact with the server using an MCP client that can connect via stdio or through a remote HTTP-like endpoint when available. The core capability is the query tool, which executes read-only SQL statements inside a read-only transaction. To explore a database, first connect to the server, then use query to retrieve metadata, sample data, or validate schema constraints. You can request table schemas, view column types, and run queries that read data without altering it. Use the results to inform prompts or analysis performed by your AI model.

Practical usage patterns include: - Discovering a table’s schema to understand which columns exist and their data types. - Running read-only queries to fetch design information or sample rows for context. - Building composite prompts that reference actual column names and types to guide model reasoning.

How to install

Prerequisites: you need either Docker or NPX to run the MCP server in a self-contained way.

Option A: Run via Docker (recommended for simple setup)

{
  "mcpServers": {
    "firebird": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "mcp/firebird",
        "firebird://host.docker.internal:3051/test_db"
      ]
    }
  }
}

Note: When running Docker on macOS, use host.docker.internal if the server runs on the host network (for example to access localhost). Replace /test_db with your actual database name and ensure your Firebird database is reachable at the provided URL. If you need credentials, you can embed them in the URL like firebird://SYSDBA:masterkey@host:3051/test_db.

Option B: Run via NPX (local, without Docker)

{
  "mcpServers": {
    "firebird": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-firebird",
        "firebird://localhost:3051/test_db"
      ]
    }
  }
}

Replace /test_db with your actual database name. If you use credentials, include them in the URL as shown in the Docker example.

Additional configuration and troubleshooting

Development setup describes container environment variables used by the Firebird container. These variables control database location, authentication behavior, and wire encryption. They are optional for local MCP usage but you may need to set them in your environment if you run the container directly.

Important container variables you may encounter include: - DATABASE: Path to the database file (for example, /firebird/data/test_db.fdb). - FIREBIRD_USER: Database user (default: SYSDBA). - FIREBIRD_PASSWORD: Database password (default: masterkey). - EnableLegacyClientAuth: Enable legacy client authentication (default: true). - EnableWireCrypt: Enable wire encryption (default: false). - CreateDb: Allow database creation (default: true).

Security and reliability notes

The server provides read-only access, so you can safely inspect schemas and query data without risking data modification. When using network access, prefer secure connections and limit access to trusted clients. If you operate locally with containers, ensure proper volume permissions and initialization scripts complete before you start querying.

Troubleshooting

Common issues and fixes: - Container startup failures: verify that the DATABASE path is correct and the Firebird data volume has the proper permissions. - Database initialization: ensure all initialization scripts (e.g., init.sql, test_db.sql) are present and permissions are correct. - Allow adequate time for the Firebird server to initialize before issuing queries.

Notes

This MCP server is designed to provide safe, read-only access to Firebird databases for large language models and other AI tools. Use the available query tool to fetch schema details and run non-destructive SQL statements as part of your data preparation and prompt engineering workflow.

Available tools

query

Execute read-only SQL queries against the connected database. Each query runs inside a READ ONLY transaction and returns the result set.