Define parameterized database tools in YAML and share them with agents through MCP or Toolbox SDKs.
Good to know
Requires a running Toolbox server, a supported source, and suitable credentials.
Get it on GitHub
Unclaimed
Are you @googleapis?Claim it, and your coding agent keeps it current with each release
Written by 
Are you @googleapis?Claim it, and your coding agent keeps it current with each release
Thanks. Marked useful.
Thanks. The report went to osfeed, and the line gets checked against the repository.
Connect AI agents and MCP clients to databases to query data, inspect schemas, and run configured database operations.
What you can do with MCP Toolbox for Databases
Query PostgreSQL and inspect its tables, active queries, and query plans.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Run configured MySQL queries and inspect tables and active queries.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Query BigQuery data and inspect dataset and table metadata.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Run SQL queries against Spanner databases.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Execute configured SQL queries against Bigtable.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Run configured MongoDB operations from database tools.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Run Redis commands through configured tools.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Show all 54
Explore Looker data using configured API tools.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Discover data assets through Dataplex catalog tools.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Manage Cloud SQL instances, users, and databases through the Admin API.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Manage AlloyDB clusters, instances, and users through API tools.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Call HTTP APIs through configured tools.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Export tool traces and metrics using OpenTelemetry.
Good to know
Requires an appropriate telemetry exporter and backend.
Wrong?
Inspect and test configured tools and toolsets in the built-in browser UI.
Good to know
Start the Toolbox server with `--ui`; the interface uses the configured tools and source permissions.
Wrong?
Run SQL queries against Oracle through configured tools.
Good to know
Requires an Oracle database and connection credentials.
Wrong?
Connect to AlloyDB and Cloud SQL databases through Private Service Connect.
Good to know
Requires a configured PSC endpoint and source `ipType: psc`.
Wrong?
Compile a local Dataform project through a tool.
Good to know
Requires the Dataform project and local compilation dependencies.
Wrong?
Retrieve, create, modify, and delete LookML project files.
Good to know
Requires Looker access; file mutations require a development-mode session.
Wrong?
Search Elasticsearch through configured tools.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Execute SQL queries through a configured MindsDB source.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Run SQL tools against SingleStore.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Retrieve FHIR and DICOM data through Cloud Healthcare tools.
Good to know
Requires Cloud Healthcare datasets and API permissions.
Wrong?
List, inspect, and cancel Serverless Spark batches.
Good to know
Requires a Google Cloud project and Serverless Spark permissions.
Wrong?
Define reusable prompts in Toolbox configuration for MCP clients.
Good to know
Requires a client supporting MCP prompts.
Wrong?
Run SQL and database inspection tools against MariaDB.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Convert natural-language database questions with Gemini Data Analytics.
Good to know
Requires the Gemini Data Analytics API, configured database context, and suitable cloud credentials.
Wrong?
Run SQL tools against Snowflake.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Generate embeddings for PostgreSQL query parameters.
Good to know
Requires an `embeddingModel` configuration and compatible embedding provider; parameters use `embeddedBy`.
Wrong?
Invoke configured tools directly from the Toolbox CLI.
Good to know
Requires the tool configuration and source access.
Wrong?
Generate Agent Skills from configured toolsets.
1 other tool for: Generate agent skillsGood to know
Generated skills require an appropriate agent runtime and access to their tools.
Wrong?
Run SQL tools against CockroachDB.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Use Cloud Logging administration tools.
Good to know
Requires Google Cloud Logging permissions.
Wrong?
Run prebuilt database tools against AlloyDB Omni.
Good to know
Requires an AlloyDB Omni instance and database credentials.
Wrong?
List and retrieve Dataproc clusters and jobs.
Good to know
Requires Google Cloud Dataproc API access.
Wrong?
Authenticate MCP access with a generic authorization service and Protected Resource Metadata.
Good to know
Requires compatible authorization configuration and client support.
Wrong?
Use vector-assist tools in Cloud SQL PostgreSQL.
Good to know
Requires Cloud SQL PostgreSQL with the relevant vector-assist support.
Wrong?
List and read Cloud Storage objects, and create, copy, move, or delete objects and manage buckets.
Good to know
Requires Google Cloud Storage credentials and permissions for the requested operations.
Wrong?
Serve Toolbox over an HTTPS/TLS listener.
Good to know
Requires a TLS certificate and private key.
Wrong?
Inspect data lineage with Google Cloud Data Lineage tools.
Good to know
Requires the Data Lineage API and appropriate cloud permissions.
Wrong?
Run Toolbox using a native Windows ARM64 binary.
Good to know
Requires Windows on ARM64.
Wrong?
Run configured tools against ScyllaDB.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Verify downloaded Toolbox binaries using their release signatures.
Good to know
Requires the signature and verification key; availability varies by release artifact.
Wrong?
Run configured queries against ArcadeDB through MCP tools.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Expose selected tools and prompts together through named MCP groups.
Good to know
Group definitions and endpoint selection determine the exposed collection.
Wrong?
Check Toolbox liveness through the `/healthz` endpoint.
Good to know
The endpoint is a container/server health check.
Wrong?
Inspect AlloyDB query performance through Advanced Query Insights tools.
Good to know
Requires Database Insights access and supported AlloyDB query-insights configuration.
Wrong?
Enable source-level read-only execution and suppress write-capable tools on supported database integrations.
Good to know
Support is source-specific; AlloyDB session read-only enforcement requires PostgreSQL 17 or newer.
Wrong?
Run Cypher queries and inspect graphs in FalkorDB.
Good to know
Requires a configured source and credentials with permission for the requested operations.
Wrong?
Execute MQL and inspect schemas for Firestore with MongoDB compatibility.
Good to know
Requires a compatible Firestore database and credentials.
Wrong?
Expose static text, files, and parameterized file resources to MCP clients.
Good to know
Requires clients supporting MCP resources; file templates need allowed paths and enforce file access limits.
Wrong?
Attach interactive HTML interfaces to tools using MCP Apps.
Good to know
Requires an MCP Apps-compatible client and UI resources with the MCP Apps MIME type.
Wrong?
Query Spanner Omni through the Spanner source.
Good to know
Set instanceType to omni and omniEndpoint; configure TLS, mTLS, or password authentication. Catalog search is unsupported.
Wrong?
Supply sensitive tool parameters out of band through the Toolbox extension.
Good to know
Experimental; requires MCP 2026-07-28 and the Toolbox extension. Secure parameters cannot have defaults, authServices, or required:false.
Wrong?
Before you upgrade MCP Toolbox for Databases
Newest firstWith `allowedDatasets`, avoid `SET @@dataset_id`, `EXTERNAL_QUERY`, region-level `INFORMATION_SCHEMA`, and unanalyzable function creation; these are blocked.
Read the evidence
Wrong?
Remove `--defer-source-connect` from startup commands; v1.13.1 removes the v1.13.0 option and restores connection initialization at startup.
Read the evidence
Wrong?
Toolbox checks its version at startup.
Read the evidence
Wrong?
Show 42 older
Allow intended internal endpoints explicitly if they use IETF protocol-assignment addresses; the default SSRF guard now blocks that range.
Read the evidence
Wrong?
Provide non-empty `type` and `id` when generating a Looker embed URL.
Read the evidence
Wrong?
AlloyDB reports an actionable error when session read-only mode is requested on PostgreSQL versions before 17.
Read the evidence
Wrong?
UUID results are strings and active-query listings exclude background processes.
Read the evidence
Wrong?
Rename prebuilt tool calls from `execute_sql_dql` to `execute_sql_readonly`.
Read the evidence
Wrong?
Fix invalid tool names or invalid tool definitions before starting the server; startup validation is enforced again.
Read the evidence
Wrong?
The default toolset is listed in alphabetical order.
Read the evidence
Wrong?
Explicitly allow intended endpoints in `100.64.0.0/10` with the HTTP source allowlist; the default guard now blocks this range.
Read the evidence
Wrong?
Active-query listing defaults to 100 results instead of 10.
Read the evidence
Wrong?
Set `audience` or `clientId` for Google auth with `mcpEnabled`; disable `--enable-api` when MCP authentication is enabled.
Read the evidence
Wrong?
Set `ALLOYDB_OMNI_PASSWORD` explicitly; the prebuilt configuration no longer defaults it to an empty value.
Read the evidence
Wrong?
Explicitly allow required internal destinations with `allowedIpRanges` or `allowPrivateNetworks`; private, loopback, and link-local requests are blocked by default.
Read the evidence
Wrong?
Build with Go 1.25.8 or newer; the module requests toolchain Go 1.26.4.
Read the evidence
Wrong?
Return a matching issuer and `active: true` in opaque-token introspection; MCP-only introspection and scope fields require `mcpEnabled`.
Read the evidence
Wrong?
Empty query result sets are represented as empty arrays instead of null.
Read the evidence
Wrong?
Update Go import and installation paths from `github.com/googleapis/genai-toolbox` to `github.com/googleapis/mcp-toolbox`.
Read the evidence
Wrong?
Provide a location for lookup-context requests; the project constraint has been relaxed.
Read the evidence
Wrong?
Move queries supplied at invocation time to `elasticsearch-execute-esql`; the configured Elasticsearch tool no longer accepts query passthrough.
Read the evidence
Wrong?
Replace `looker-git-branch` with `looker-get-git-branch`, `looker-list-git-branches`, `looker-create-git-branch`, `looker-delete-git-branch`, or `looker-switch-git-branch`.
Read the evidence
Wrong?
Use `/mcp`, or explicitly enable the legacy `/api` endpoint with `--enable-api`; it is now disabled by default.
Read the evidence
Wrong?
Replace `--tools_file` with `--config`; migrate deprecated `--tools-file`, `--tools-files`, and `--tools-folder` to `--config`, `--configs`, and `--config-folder`.
Read the evidence
Wrong?
Use singular flat-format kinds such as `source` and `tool`; replace `authSources` with supported auth-service configuration and remove Cloud SQL SQL Server `ipAddress`.
Read the evidence
Wrong?
Non-2xx errors return sanitized output instead of the original upstream error content.
Read the evidence
Wrong?
Update named prebuilt toolset references to the new task groups; for PostgreSQL, replace `postgres_database_tools` with `data`, `monitor`, `health`, `view-config`, or `replication`.
Read the evidence
Wrong?
Replace old transport/API metrics with `mcp.server.operation.duration`, `mcp.server.session.duration`, `toolbox.server.mcp.active_sessions`, and `toolbox.tool.operation.duration`.
Read the evidence
Wrong?
Build with Go 1.25.7 or newer; the module now requests toolchain Go 1.25.8.
Read the evidence
Wrong?
Generate one skill per toolset when no individual toolset is selected.
Read the evidence
Wrong?
Supply non-null values for required parameters; explicit nulls now fail required-value validation.
Read the evidence
Wrong?
Configuration supports flat YAML documents separating `kind`, `name`, and `type`; legacy nested configuration is still converted automatically.
Read the evidence
Wrong?
MCP traces adopt OpenTelemetry semantic conventions; review trace queries that rely on older attribute names.
Read the evidence
Wrong?
Rename invalid tools to use only letters, digits, underscores, hyphens, and periods; names must satisfy the new MCP naming validation.
Read the evidence
Wrong?
Send the natural-language input as `query` instead of `prompt`.
Read the evidence
Wrong?
Numeric query results are formatted as decimal strings.
Read the evidence
Wrong?
MSSQL sources now supply default host and port values when omitted.
Read the evidence
Wrong?
Rename view input `viewname` to `view_name`; sequence fields `sequencename`, `schemaname`, and output `sequenceowner` become `sequence_name`, `schema_name`, and `sequence_owner`.
Read the evidence
Wrong?
Schema listing now defaults to a maximum of 50 rows; set `limit` to request a different bound.
Read the evidence
Wrong?
Update response parsing for the added Cloud Console and Logging URLs in batch creation, listing, and retrieval outputs.
Read the evidence
Wrong?
Read `object_details` as nested JSON rather than a JSON-encoded string in list-tables responses.
Read the evidence
Wrong?
Review `allowedDatasets` before using dataset-info or contribution analysis; these tools now enforce the source restriction.
Read the evidence
Wrong?
Recreate old AlloyDB AI NL configurations using `alloydb_ai_nl.g_create_configuration` instead of the old create-configuration operation.
Read the evidence
Wrong?
PostgreSQL and MySQL sources now supply default host and port values when omitted.
Read the evidence
Wrong?
Review `allowedDatasets`: table-info now enforces it; list-dataset-ids returns the configured datasets and ignores `project` when restrictions are set.
Read the evidence
Wrong?
MCP Toolbox for Databases release history 26
Connect database tools to Spanner Omni.
Details
Adds Spanner Omni connections and supports authorized BigQuery views under dataset restrictions.
Also: Looker gains Explore retrieval and LookML dashboard discovery.
v1.14.0; use the release binaries or container image and configure the required source credentials.
Added: Query Spanner Omni through the Spanner source.Improved: Query BigQuery data and inspect dataset and table m…Action required / inferred: Query BigQuery data and inspect dataset and table m…Improved: Explore Looker data using configured API tools.Improved: Convert natural-language database questions with Ge…Now works: Expose static text, files, and parameterized file r…
Seeded by @iluxav with Codex
v1.13.1
with v1.13.0
with v1.13.0
Load prebuilt Bigtable tools.
Details
Adds Bigtable prebuilt tools and complete or paged data-agent discovery. Reverts the source-connection deferral introduced in v1.13.0.
Also: Data-agent listing can fetch all pages; v1.13.1 removes the briefly available deferred-connection flag.
v1.13.1; use the release binaries or container image and configure the required source credentials.
Improved: Execute configured SQL queries against Bigtable.Improved: Convert natural-language database questions with Ge…Improved: Explore Looker data using configured API tools.Action required / inferred: Define parameterized database tools in YAML and sha…
Seeded by @iluxav with Codex
v1.12.0
with v1.11.0
with v1.11.0
Expose files and text as MCP resources.
Details
Adds MCP resources and interactive tool interfaces, and exposes group discovery. Adds a configurable app-verification endpoint and stricter Looker embed-URL inputs.
Also: Also adds MCP Apps, app verification, and stricter Looker embed-URL inputs.
v1.12.0; use the release binaries or container image and configure the required source credentials.
Added: Expose static text, files, and parameterized file r…Added: Attach interactive HTML interfaces to tools using M…Improved: Expose selected tools and prompts together through…Improved: Export tool traces and metrics using OpenTelemetry.Behavior changed: Define parameterized database tools in YAML and sha…Action required: Call HTTP APIs through configured tools.Improved: Attach interactive HTML interfaces to tools using M…Improved: Explore Looker data using configured API tools.Action required / inferred: Explore Looker data using configured API tools.Behavior changed: Enable source-level read-only execution and suppres…
Seeded by @iluxav with Codex
Enforce native read-only execution for supported database sources.
Details
Adds source-level read-only controls, FalkorDB tools, and an experimental secure-parameter extension.
v1.10.0; use the release binaries or container image and configure the required source credentials.
Added: Enable source-level read-only execution and suppres…Added: Run Cypher queries and inspect graphs in FalkorDB.Added: Execute MQL and inspect schemas for Firestore with…Added: Supply sensitive tool parameters out of band throug…Improved: Run configured MongoDB operations from database too…Now works: List and read Cloud Storage objects, and create, co…Now works: Define parameterized database tools in YAML and sha…Behavior changed: Query PostgreSQL and inspect its tables, active que…
Seeded by @iluxav with Codex
Read the evidence
Inspect AlloyDB performance with Advanced Query Insights tools.
Details
Adds AlloyDB query insights and Bigtable administration, and renames Spanner's read-only SQL tool.
v1.9.0; use the release binaries or container image and configure the required source credentials.
Added: Inspect AlloyDB query performance through Advanced…Improved: Execute configured SQL queries against Bigtable.Improved: Manage Cloud SQL instances, users, and databases th…Improved: Explore Looker data using configured API tools.Improved: Expose selected tools and prompts together through…Improved: Query BigQuery data and inspect dataset and table m…Action required: Run SQL queries against Spanner databases.Now works: Authenticate MCP access with a generic authorizatio…Now works: Define parameterized database tools in YAML and sha…
Seeded by @iluxav with Codex
Read the evidence
Organize tools and prompts into named groups.
Details
Adds named groups, group-based skill generation, and a server health endpoint.
Also: Startup validation and HTTP address restrictions become stricter.
v1.8.0; use the release binaries or container image and configure the required source credentials.
Added: Expose selected tools and prompts together through…Improved: Generate Agent Skills from configured toolsets.Added: Check Toolbox liveness through the `/healthz` endpo…Improved: Explore Looker data using configured API tools.Improved: Discover data assets through Dataplex catalog tools.Action required: Define parameterized database tools in YAML and sha…Behavior changed: Define parameterized database tools in YAML and sha…Action required: Call HTTP APIs through configured tools.Behavior changed / inferred: Run configured MySQL queries and inspect tables and…
Seeded by @iluxav with Codex
Run database tools against ArcadeDB.
Details
Adds ArcadeDB tools, ClickHouse embedding parameters, and catalog data-product management.
v1.7.0; use the release binaries or container image and configure the required source credentials.
Added: Run configured queries against ArcadeDB through MCP…Improved: Generate embeddings for PostgreSQL query parameters.Improved: Discover data assets through Dataplex catalog tools.Improved: Convert natural-language database questions with Ge…Improved: Query PostgreSQL and inspect its tables, active que…Now works: Explore Looker data using configured API tools.
Seeded by @iluxav with Codex
Read the evidence
Use calculated fields and complex filters in Looker queries.
Details
Adds richer Looker query parameters and catalog data-product tools; release binaries gain signatures.
v1.6.0; use the release binaries or container image and configure the required source credentials.
Improved: Explore Looker data using configured API tools.Improved: List and read Cloud Storage objects, and create, co…Improved: Discover data assets through Dataplex catalog tools.Improved: Define parameterized database tools in YAML and sha…Now works: Convert natural-language database questions with Ge…Now works: Retrieve, create, modify, and delete LookML project…Added: Verify downloaded Toolbox binaries using their rele…
Seeded by @iluxav with Codex
Read the evidence
Generate agent skills without connecting to live databases.
Details
Enables offline skill generation, source-level access restrictions, and selected prebuilt toolsets.
Also: MCP auth, internal HTTP destinations, and AlloyDB Omni password configuration may require changes.
v1.5.0; use the release binaries or container image and configure the required source credentials.
Improved: Generate Agent Skills from configured toolsets.Added: Run configured tools against ScyllaDB.Improved: Define parameterized database tools in YAML and sha…Improved: List and read Cloud Storage objects, and create, co…Improved: Run configured MySQL queries and inspect tables and…Improved: Discover data assets through Dataplex catalog tools.Action required: Authenticate MCP access with a generic authorizatio…Action required / inferred: Run prebuilt database tools against AlloyDB Omni.Action required: Call HTTP APIs through configured tools.Action required / inferredImproved: Export tool traces and metrics using OpenTelemetry.Now works: Query BigQuery data and inspect dataset and table m…Now works: Run configured MySQL queries and inspect tables and…
Seeded by @iluxav with Codex
Trace data lineage through configured tools.
Details
Adds Data Lineage tools and Windows ARM64 support; generic token validation is stricter.
v1.4.0; use the release binaries or container image and configure the required source credentials.
Added: Inspect data lineage with Google Cloud Data Lineage…Added: Run Toolbox using a native Windows ARM64 binary.Improved: Define parameterized database tools in YAML and sha…Improved: Use vector-assist tools in Cloud SQL PostgreSQL.Improved: Run SQL queries against Spanner databases.Action required / inferred: Authenticate MCP access with a generic authorizatio…Now works: Query BigQuery data and inspect dataset and table m…Now works: Run SQL queries against Oracle through configured t…
Seeded by @iluxav with Codex
Restrict individual tools by OAuth scope.
Details
Adds tool-level authorization scopes and multi-database Cloud SQL execution; toolset boundaries are enforced.
v1.3.0; use the release binaries or container image and configure the required source credentials.
Improved: Authenticate MCP access with a generic authorizatio…Now works: Define parameterized database tools in YAML and sha…Now works: Call HTTP APIs through configured tools.Improved: Manage Cloud SQL instances, users, and databases th…Improved: Export tool traces and metrics using OpenTelemetry.Behavior changed: Define parameterized database tools in YAML and sha…Now works: Explore Looker data using configured API tools.
Seeded by @iluxav with Codex
Read the evidence
Manage Cloud Storage objects through tools.
Details
Adds Cloud Storage operations, a TLS listener, and BigQuery query-cost limits.
v1.2.0; use the release binaries or container image and configure the required source credentials.
Added: List and read Cloud Storage objects, and create, co…Added: Serve Toolbox over an HTTPS/TLS listener.Improved: Query BigQuery data and inspect dataset and table m…Improved: Discover data assets through Dataplex catalog tools.Now works: Explore Looker data using configured API tools.Now works: Define parameterized database tools in YAML and sha…Now works: Query BigQuery data and inspect dataset and table m…
Seeded by @iluxav with Codex
Read the evidence
v1.1.0
with v0.32.0, v1.0.0
with v0.32.0, v1.0.0
Use database tools under the project's first stable API release.
Details
Covers v0.32.0, the first stable v1.0.0, and v1.1.0: updates the Go module name, adds data-agent and vector-search tools, and separates Looker Git-branch operations.
Also: Adds Cloud SQL PostgreSQL vector assist and repairs Looker configuration.
v1.1.0; use the release binaries or container image and configure the required source credentials.
Improved: Convert natural-language database questions with Ge…Improved: Explore Looker data using configured API tools.Improved: Generate Agent Skills from configured toolsets.Action required / inferred: Define parameterized database tools in YAML and sha…Action required: Discover data assets through Dataplex catalog tools.Improved: Run SQL tools against SingleStore.Now works: Query BigQuery data and inspect dataset and table m…Now stable: Define parameterized database tools in YAML and sha…Now stable: Query PostgreSQL and inspect its tables, active que…Now stable: Run configured MySQL queries and inspect tables and…Now stable: Query BigQuery data and inspect dataset and table m…Now stable: Run SQL queries against Spanner databases.Now stable: Execute configured SQL queries against Bigtable.Now stable: Run configured MongoDB operations from database too…Now stable: Run Redis commands through configured tools.Now stable: Explore Looker data using configured API tools.Now stable: Discover data assets through Dataplex catalog tools.Now stable: Manage Cloud SQL instances, users, and databases th…Now stable: Manage AlloyDB clusters, instances, and users throu…Now stable: Call HTTP APIs through configured tools.Now stable: Export tool traces and metrics using OpenTelemetry.Now stable: Run SQL queries against Oracle through configured t…Now stable: Connect to AlloyDB and Cloud SQL databases through…Now stable: Compile a local Dataform project through a tool.Now stable: Retrieve, create, modify, and delete LookML project…Now stable: Search Elasticsearch through configured tools.Now stable: Execute SQL queries through a configured MindsDB so…Now stable: Run SQL tools against SingleStore.Now stable: Retrieve FHIR and DICOM data through Cloud Healthca…Now stable: List, inspect, and cancel Serverless Spark batches.Now stable: Define reusable prompts in Toolbox configuration fo…Now stable: Run SQL and database inspection tools against Maria…Now stable: Convert natural-language database questions with Ge…Now stable: Run SQL tools against Snowflake.Now stable: Generate embeddings for PostgreSQL query parameters.Now stable: Invoke configured tools directly from the Toolbox C…Now stable: Generate Agent Skills from configured toolsets.Now stable: Run SQL tools against CockroachDB.Now stable: Use Cloud Logging administration tools.Now stable: Run prebuilt database tools against AlloyDB Omni.Now stable: List and retrieve Dataproc clusters and jobs.Now stable: Authenticate MCP access with a generic authorizatio…Improved: Generate embeddings for PostgreSQL query parameters.Improved: Search Elasticsearch through configured tools.Action required: Search Elasticsearch through configured tools.Action required: Retrieve, create, modify, and delete LookML project…Improved: Authenticate MCP access with a generic authorizatio…Improved: Run configured MySQL queries and inspect tables and…Added: Use vector-assist tools in Cloud SQL PostgreSQL.Now works: Explore Looker data using configured API tools.Now stable: Inspect and test configured tools and toolsets in t…Limit lifted: Define parameterized database tools in YAML and sha…
Seeded by @iluxav with Codex
Authenticate MCP clients with a generic authorization service.
Details
Adds generic MCP authorization and disables the legacy API by default; configuration names and CLI flags change.
v0.31.0; use the release binaries or container image and configure the required source credentials. Beta release; the public API can change.
Added: Authenticate MCP access with a generic authorizatio…Action required: Define parameterized database tools in YAML and sha…Action required: Define parameterized database tools in YAML and sha…Action required: Define parameterized database tools in YAML and sha…Behavior changed: Call HTTP APIs through configured tools.Improved: Discover data assets through Dataplex catalog tools.Now works: Generate Agent Skills from configured toolsets.
Seeded by @iluxav with Codex
Read the evidence
v0.30.0
with v0.29.0
with v0.29.0
Convert existing configuration files with a migration command.
Details
Splits prebuilt toolsets by task and replaces the telemetry metric surface. Adds configuration migration and explicit server startup commands, and generates separate skills per toolset.
Also: Prebuilt toolsets and metric names change; source builds require newer Go.
v0.30.0; use the release binaries or container image and configure the required source credentials. Beta release; the public API can change.
Action required / inferred: Define parameterized database tools in YAML and sha…Action required: Export tool traces and metrics using OpenTelemetry.Action required / inferredImproved: Query BigQuery data and inspect dataset and table m…Improved: List, inspect, and cancel Serverless Spark batches.Now works: Generate Agent Skills from configured toolsets.Now works: Run SQL queries against Oracle through configured t…Improved: Define parameterized database tools in YAML and sha…Behavior changed: Generate Agent Skills from configured toolsets.Improved: Retrieve, create, modify, and delete LookML project…Improved: Discover data assets through Dataplex catalog tools.Now works: Run SQL queries against Oracle through configured t…Now works: Query PostgreSQL and inspect its tables, active que…
Seeded by @iluxav with Codex
Inspect Dataproc clusters and jobs through tools.
Details
Adds Dataproc inspection, Redis TLS, and more LookML development tools.
v0.28.0; use the release binaries or container image and configure the required source credentials. Beta release; the public API can change.
Added: List and retrieve Dataproc clusters and jobs.Improved: Run Redis commands through configured tools.Improved: Query PostgreSQL and inspect its tables, active que…Improved: Retrieve, create, modify, and delete LookML project…Now works: Run SQL queries against Oracle through configured t…Action required: Define parameterized database tools in YAML and sha…Improved: Define parameterized database tools in YAML and sha…
Seeded by @iluxav with Codex
Read the evidence
Invoke tools directly from the command line.
Details
Adds direct CLI invocation and skill generation, with new database and logging integrations.
Also: Flat configuration is available while nested files remain compatible.
v0.27.0; use the release binaries or container image and configure the required source credentials. Beta release; the public API can change.
Added: Invoke configured tools directly from the Toolbox C…Added: Generate Agent Skills from configured toolsets.Added: Run SQL tools against CockroachDB.Added: Use Cloud Logging administration tools.Added: Run prebuilt database tools against AlloyDB Omni.Behavior changed: Define parameterized database tools in YAML and sha…Behavior changed: Export tool traces and metrics using OpenTelemetry.Improved: Retrieve, create, modify, and delete LookML project…Now works: Discover data assets through Dataplex catalog tools.
Seeded by @iluxav with Codex
Read the evidence
Back up and restore Cloud SQL instances through tools.
Details
Adds Cloud SQL backup tools and multiple prebuilt configurations; tool names and GDA input names require migration.
v0.26.0; use the release binaries or container image and configure the required source credentials. Beta release; the public API can change.
Improved: Manage Cloud SQL instances, users, and databases th…Action required: Define parameterized database tools in YAML and sha…Action required: Convert natural-language database questions with Ge…Improved: Define parameterized database tools in YAML and sha…Improved: Query BigQuery data and inspect dataset and table m…Improved: Generate embeddings for PostgreSQL query parameters.
Seeded by @iluxav with Codex
Read the evidence
Query Snowflake through configured tools.
Details
Adds Snowflake tools and embedding models for PostgreSQL parameters.
Also: Cloud SQL MySQL prebuilt configurations now permit IAM without user/password variables.
v0.25.0; use the release binaries or container image and configure the required source credentials. Beta release; the public API can change.
Added: Run SQL tools against Snowflake.Added: Generate embeddings for PostgreSQL query parameters.Improved: Define parameterized database tools in YAML and sha…Improved / inferred: Run configured MySQL queries and inspect tables and…Improved: Query BigQuery data and inspect dataset and table m…Improved: Explore Looker data using configured API tools.Improved: Query PostgreSQL and inspect its tables, active que…Now works: Convert natural-language database questions with Ge…
Seeded by @iluxav with Codex
Generate database queries with Gemini Data Analytics.
Details
Adds Gemini Data Analytics integration and permits combined prebuilt and custom configurations.
v0.24.0; use the release binaries or container image and configure the required source credentials. Beta release; the public API can change.
Added: Convert natural-language database questions with Ge…Improved: Run SQL queries against Oracle through configured t…Improved: Define parameterized database tools in YAML and sha…Improved: Run configured MySQL queries and inspect tables and…Now works: Run SQL queries against Spanner databases.
Seeded by @iluxav with Codex
Read the evidence
v0.23.0
with v0.22.0
with v0.22.0
Run MySQL-compatible tools against MariaDB.
Details
Adds Spark batch creation and database inspection tools; BigQuery numeric output becomes decimal strings. Adds MariaDB support and PostgreSQL statistics; Spark responses gain console and logging links.
Also: Also adds Spark batch creation, PostgreSQL statistics, and new output formats.
v0.23.0; use the release binaries or container image and configure the required source credentials. Beta release; the public API can change.
Improved: List, inspect, and cancel Serverless Spark batches.Improved: Query PostgreSQL and inspect its tables, active que…Improved: Run SQL queries against Spanner databases.Improved: Manage Cloud SQL instances, users, and databases th…Improved: Define parameterized database tools in YAML and sha…Behavior changed: Query BigQuery data and inspect dataset and table m…Behavior changed: Define parameterized database tools in YAML and sha…Now works: Define parameterized database tools in YAML and sha…Added: Run SQL and database inspection tools against Maria…Improved: Query PostgreSQL and inspect its tables, active que…Action required / inferred: Query PostgreSQL and inspect its tables, active que…Behavior changed / inferred: Query PostgreSQL and inspect its tables, active que…Action required: List, inspect, and cancel Serverless Spark batches.Improved: Explore Looker data using configured API tools.
Seeded by @iluxav with Codex
Inspect PostgreSQL locks and long-running transactions.
Details
Adds PostgreSQL operational tools and changes Spanner table metadata to nested JSON.
v0.21.0; use the release binaries or container image and configure the required source credentials. Beta release; the public API can change.
Improved: Query PostgreSQL and inspect its tables, active que…Action required: Run SQL queries against Spanner databases.Now works: Query PostgreSQL and inspect its tables, active que…
Seeded by @iluxav with Codex
Read the evidence
v0.20.0
with v0.19.1
with v0.19.1
Serve reusable prompts to MCP clients.
Details
Adds database integrations, Spark batch inspection, and broader schema inspection. Adds configurable prompts and PostgreSQL inspection tools.
Also: Also adds Elasticsearch, MindsDB, SingleStore, Cloud Healthcare, and Spark tools, with BigQuery and AlloyDB upgrade notes.
v0.20.0; use the release binaries or container image and configure the required source credentials. Beta release; the public API can change.
Added: Search Elasticsearch through configured tools.Added: Execute SQL queries through a configured MindsDB so…Added: Run SQL tools against SingleStore.Added: Retrieve FHIR and DICOM data through Cloud Healthca…Added: List, inspect, and cancel Serverless Spark batches.Improved: Query PostgreSQL and inspect its tables, active que…Improved: Explore Looker data using configured API tools.Improved: Run configured MongoDB operations from database too…Improved: Query BigQuery data and inspect dataset and table m…Action required: Query BigQuery data and inspect dataset and table m…Action required: Define parameterized database tools in YAML and sha…Improved: Define parameterized database tools in YAML and sha…Added: Define reusable prompts in Toolbox configuration fo…Improved: Query PostgreSQL and inspect its tables, active que…Improved: Explore Looker data using configured API tools.Behavior changed: Define parameterized database tools in YAML and sha…Improved: Manage Cloud SQL instances, users, and databases th…Improved: Manage AlloyDB clusters, instances, and users throu…
Seeded by @iluxav with Codex
Create and edit LookML project files through tools.
Details
Adds LookML file tools and parameter constraints; HTTP tools accept the full 2xx success range.
v0.18.0; use the release binaries or container image and configure the required source credentials. Beta release; the public API can change.
Added: Retrieve, create, modify, and delete LookML project…Improved: Define parameterized database tools in YAML and sha…Now works: Call HTTP APIs through configured tools.
Seeded by @iluxav with Codex
Read the evidence
Run configured SQL tools against Oracle databases.
Details
Adds Oracle tools and PSC connections, and extends Looker and BigQuery operations.
Also: Also adds local Dataform compilation and fixes MySQL table listing.
v0.17.0; use the release binaries or container image and configure the required source credentials. Beta release; the public API can change.
Added: Run SQL queries against Oracle through configured t…Improved: Query BigQuery data and inspect dataset and table m…Action required: Query BigQuery data and inspect dataset and table m…Added: Connect to AlloyDB and Cloud SQL databases through…Added: Compile a local Dataform project through a tool.Improved: Explore Looker data using configured API tools.Now works: Run configured MySQL queries and inspect tables and…
Seeded by @iluxav with Codex
Read the evidence
v0.16.0
Starting point
Starting point
Connect agents to configured database queries through MCP.
Details
YAML defines sources, parameterized tools, and named toolsets. SDKs load these tools for applications, and configuration reloads dynamically by default.
v0.16.0; use the release binaries or container image and configure the required source credentials.
Added: Define parameterized database tools in YAML and sha…Added: Query PostgreSQL and inspect its tables, active que…Added: Run configured MySQL queries and inspect tables and…Added: Query BigQuery data and inspect dataset and table m…Added: Run SQL queries against Spanner databases.Added: Execute configured SQL queries against Bigtable.Added: Run configured MongoDB operations from database too…Added: Run Redis commands through configured tools.Added: Explore Looker data using configured API tools.Added: Discover data assets through Dataplex catalog tools.Added: Manage Cloud SQL instances, users, and databases th…Added: Manage AlloyDB clusters, instances, and users throu…Added: Call HTTP APIs through configured tools.Added: Export tool traces and metrics using OpenTelemetry.Added: Inspect and test configured tools and toolsets in t…
Seeded by @iluxav with Codex
Read the evidence
- googleapis/mcp-toolbox/blob/v0.16.0/README.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/resources/sources/postgres.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/resources/sources/mysql.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/resources/sources/bigquery.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/resources/sources/spanner.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/resources/sources/bigtable.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/resources/sources/mongodb.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/resources/sources/redis.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/resources/sources/looker.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/resources/sources/dataplex.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/resources/sources/cloud-sql-admin.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/resources/sources/alloydb-admin.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/resources/sources/http.md
- googleapis/mcp-toolbox/blob/v0.16.0/docs/en/how-to/toolbox-ui/index.md