Multi-Connection

한국어: 멀티커넥션

One server process can serve several CUBRID databases. Every tool accepts an optional connection argument selecting the target; single-database setups are unchanged.

Default connection

The bare CUBRID_* variables define a single connection named default:

export CUBRID_HOST=localhost
export CUBRID_USER=readonly_user
export CUBRID_PASSWORD=secret
export CUBRID_DATABASE=mydb

Named connections

List extra connection names in CUBRID_CONNECTIONS (comma-separated) and provide CUBRID_<NAME>_* variables for each:

# Default connection (unchanged)
export CUBRID_HOST=localhost
export CUBRID_USER=readonly_user
export CUBRID_PASSWORD=secret
export CUBRID_DATABASE=mydb

# Additional named connections
export CUBRID_CONNECTIONS=reporting,analytics

export CUBRID_REPORTING_HOST=reporting-db
export CUBRID_REPORTING_USER=readonly_user
export CUBRID_REPORTING_PASSWORD=secret
export CUBRID_REPORTING_DATABASE=reports
export CUBRID_REPORTING_MCP_MAX_ROWS=500   # optional per-connection tuning

export CUBRID_ANALYTICS_HOST=analytics-db
export CUBRID_ANALYTICS_USER=readonly_user
export CUBRID_ANALYTICS_PASSWORD=secret
export CUBRID_ANALYTICS_DATABASE=analytics

Then, in a client conversation: "Query the analytics database for …" — the model passes connection="analytics" to the tool.

Rules

• Connection names must match [A-Za-z0-9_]+ and are matched case-insensitively.
• default is reserved (it always comes from the bare CUBRID_* variables) and cannot appear in CUBRID_CONNECTIONS.
• For a named connection <NAME>, connection fields live at CUBRID_<NAME>_HOST etc., and the optional tuning knobs at CUBRID_<NAME>_MCP_* (the same suffixes as the global ones).
• Named connections do not inherit values from the bare variables — specify every field for each named connection.
• Selecting an unknown connection returns a clear error listing the available names.

Per-connection isolation

Each connection is independently configured and enforced:

Setting	Per-connection variable	Effect
Read-only	CUBRID_<NAME>_MCP_READONLY	Whitelist enforcement for this connection only
Write mode	CUBRID_<NAME>_MCP_WRITE	Opt-in execute_write for this connection only
Audit log	CUBRID_<NAME>_MCP_AUDIT_LOG	Audit stream for this connection only
Output limits	CUBRID_<NAME>_MCP_MAX_ROWS / _MAX_CHARS / _MAX_SQL_LENGTH / _QUERY_TIMEOUT / _CONNECT_TIMEOUT / _MAX_ROW_COUNT_TABLES / _MAX_INLINE_BINARY_BYTES	Tuning for this connection only

Each connection also has its own lock, connection lifecycle, and stale-connection recovery, so a hung query on one database does not block the others.

Write-mode registration nuance: the execute_write tool appears in MCP capability discovery when any connection enables writes, but every call is enforced against the target connection's setting — a connection with writes off refuses the write even when another connection enables them. See the Security Model.

Resources use the default connection

Every tool accepts the connection argument, but the MCP resources do not: cubrid://schema and cubrid://schema/{table} always read the default connection, because a resource URI carries no connection selector. To inspect a named connection's schema, call the tools (all_table_names, describe_table, …) with connection="<name>".