Enables AI agents to interact with Google BigQuery databases, including listing tables, retrieving schemas, and executing SQL queries.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@BigQuery MCP What are the top 10 most popular names in 2020?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A Model Context Protocol (MCP) server for Google BigQuery that enables AI agents to interact with BigQuery databases through natural language queries and schema exploration.
This project was developed by Snow Leopard AI as a benchmarking tool for our platform, and we're making it publicly available for the community to use and build upon.
The Model Context Protocol (MCP) is an open standard that allows AI applications to securely connect to external data sources and tools. This BigQuery MCP server acts as a bridge between AI agents and your BigQuery datasets.
Related MCP server: BigQuery MCP Server
Resource URI | Description |
| List all tables available to the agent |
| Get the schema of a specific table |
Tool | Description |
| List available tables |
| Get the schema of a given table |
| Execute BigQuery SQL and return results |
Before getting started, ensure you have:
Claude Desktop: Download here
Google Cloud Project with BigQuery enabled: Setup guide
Google Cloud CLI (gcloud): Installation guide
UV Package Manager: Installation guide
First, we need to authenticate with Google.
gcloud auth application-default loginThis opens your browser to authenticate your local machine with Google Cloud.
Edit your claude_desktop_config.json file to add the BigQuery MCP server.
Application: Claude > Settings > Developer > Edit Config
Mac: ~/Library/Application\ Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\\Claude\\claude_desktop_config.json
You will need to set your project to a Google Cloud project with permissions to submit bigquery jobs. If you do not have a project that you can run bigquery jobs on, create and test one by following Google's BigQuery Quickstart Guide Create a project and follow the instructions to query a public dataset.
{
"mcpServers": {
"bigquery": {
"command": "uvx",
"args": [
"sl-bigquery-mcp",
"--dataset",
"bigquery-public-data.usa_names",
"--project",
"🚨 <projectName> 🚨"
]
}
}
}Depending on how you have installed uv, the uvx executable may not be in Claude Desktop's PATH if it is launched from the GUI. To be sure uvx is accessible from Claude Desktop, let's run it in the terminal.
open -a claudeAfter saving the configuration, restart Claude Desktop. You should now be able to ask Claude questions about your BigQuery data!
What are the top 10 most popular names in 2020?To see a complete list of parameters:
uvx sl-bigquery-mcp --helpUsage: sl-bigquery-mcp [OPTIONS]
╭─ Options ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ --mode [stdio|sse|streamable-http] MCP transport protocol [default: stdio] │
│ --dataset TEXT Dataset(s) for mcp resources. Will create resources for all tables. │
│ --table TEXT Table(s) for mcp resources. Can be specified as project.dataset.table or dataset.table │
│ --enable-list-tables-tool --no-enable-list-tables-tool Registers list_resources tool [default: enable-list-tables-tool] │
│ --enable-schema-tool --no-enable-schema-tool Registers get_schema tool [default: enable-schema-tool] │
│ --project TEXT BigQuery project [env var: BQ_PROJECT] [default: None] │
│ --api-method [INSERT|QUERY] BigQuery client api_method [default: QUERY] │
│ --port INTEGER [default: 8000] │
╰───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯First, check out your Claude Desktop app logs (in the same directory as the config file) for more verbose errors / logging
This usually means Claude is having issues starting the mcp server. Frequently this is due to uvx being inaccessible from
the application. In this case, use the full path to your uvx executable instead of just uvx in claude_desktop_config.json.
To find your uv executable, run
which uvxOtherwise, this may be caused by bad arguments, dependency version incompatibilities, or bugs. If you run into the last two, please file an issue describing the problem.
This may be a misconfiguration mcp server, authentication issues, the llm getting too much data, or of course, product bugs. After checking the logs, consider using the MCP Inspector to debug your issue. And of course, file any bugs you find on our issue board.
Clone the repository
Setup virtual environment and install dependencies
Verify installation
git clone https://github.com/SnowLeopard-AI/bigquery-mcp.git
cd bigquery-mcp
uv sync
source .venv/bin/activate
sl-bigquery-mcp --helpThe following command will launch a browser for you to login to your google cloud account. You must have a Google Cloud
project with BigQuery enabled. If you don't, see Google's bigquery setup guide.
gcloud auth application-default login
gcloud config set project <projectName>
gcloud auth application-default set-quota-project <projectName>Run the tests to make sure your dev environment is properly configured.
pytest testsNote: the tests run actual BigQuery queries against public datasets and require authentication.
For hands-on testing and development, use the MCP Inspector tool:
npx @modelcontextprotocol/inspector uv run sl-bigquery-mcp --dataset bigquery-public-data.usa_namesWe welcome contributions! Please coordinate with us on discord to ensure your changes can quicly make it into the repo. Communicating before coding always saves time.
For logistics of contributing to an open source project, see the first contributions repository.
Issues: GitHub Issues
Documentation: BigQuery Documentation
MCP Protocol: Model Context Protocol
Contact: Discord Server
Get the schema for a given table
| Name | Required | Description | Default |
|---|---|---|---|
| table | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations exist, so the description carries the full burden of behavioral disclosure. It implies a read-only operation, but it does not explain what the schema includes, what happens for unknown tables, whether access is required, or what the response format will be.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single declarative sentence with no filler, and the core purpose is front-loaded. It is appropriately brief for a simple tool, though the brevity comes at the cost of missing behavioral and parameter context.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a one-parameter read-only tool, this is minimally viable: it states the operation and target. However, it omits return expectations and useful context such as linking to list_tables for discovering valid table names, leaving notable gaps for an agent to call it confidently.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%, so the description needed to clarify the 'table' parameter. It only restates the concept of 'a given table' in English and adds no details about valid values, naming conventions, or how to discover available tables.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states a specific action ('Get') and resource ('the schema for a given table'), which distinguishes it in kind from siblings like query and list_tables. It does not explicitly contrast with siblings, but the resource type is evident enough.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
There is no explicit guidance on when to use this tool versus query or list_tables. The intended use is implied by the tool name and one-line description, but no prerequisites, exclusions, or alternative conditions are provided.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List the tables available
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the full burden of behavioral disclosure. It only says 'List the tables available' — no mention of whether system tables are included, the return format, pagination, or authentication requirements. The word 'list' implies read-only, but scope is vague.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
A single, front-loaded sentence with no filler. Every word contributes to the purpose, making it appropriately concise for a zero-parameter tool.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple list tool with no parameters or output schema, this is minimally adequate. However, it doesn't clarify what 'available' means (e.g., current schema, user permissions), the return shape, or how it relates to siblings — leaving a small but notable gap.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool has zero parameters, so the schema fully documents the input surface. The description doesn't need to add parameter semantics; baseline of 4 applies.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states a specific verb ('List') and resource ('tables'), and the phrase 'available' conveys scope. Even without naming siblings, the operation is obviously distinct from 'query' (data retrieval) and 'get_schema' (schema inspection).
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description gives no guidance on when to use this tool versus alternatives. There is no mention of prerequisites, whether to call this before query/get_schema, or any context indicating selection criteria.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Executes the provided BigQuery sql statement and returns the results
| Name | Required | Description | Default |
|---|---|---|---|
| sql | Yes | BigQuery sql statement to execute |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden of disclosing behavior. It states that SQL is executed and results are returned, but it does not say whether this is read-only, whether DDL/DML is permitted, whether side effects can occur, or what happens if the query is expensive or large. This is a significant gap for an arbitrary SQL execution tool.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, direct sentence that front-loads the action and result. There is no wasted wording or redundant detail.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
With no annotations and no output schema, the description needs to provide enough context for an agent to invoke the tool safely and correctly. For an arbitrary SQL execution tool, it omits important constraints: whether only SELECT is allowed, how results are returned, row limits, and potential cost or side-effect warnings. The description is minimally functional but not complete for this kind of operation.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema already documents the single 'sql' parameter with 100% coverage. The description essentially restates the schema field ('BigQuery sql statement'), adding little meaning beyond what the structured definition provides. Baseline 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly identifies the action ('Executes'), the resource ('BigQuery sql statement'), and the outcome ('returns the results'). It is unambiguous, though it does not explicitly differentiate itself from sibling tools list_tables and get_schema, which are also BigQuery-related.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance is given about when to use this tool versus list_tables or get_schema. An agent is not told that query should be used for arbitrary SQL while the siblings cover metadata or schema access, nor are any exclusions or prerequisites stated.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.
v0.1.9get_schemalist_tablesqueryEach tool has a distinct and clear purpose: query executes SQL, list_tables enumerates tables, and get_schema describes table structure. There is little risk of an agent selecting the wrong tool for a task.
list_tables and get_schema follow a consistent verb_noun snake_case pattern. The lone tool 'query' is a simple verb without a noun object, which is a minor deviation but still clear and natural for the operation.
Three tools is on the minimal side but reasonable for a focused BigQuery read/query server. Each tool serves a distinct need, though the server could support a slightly broader set without feeling bloated.
The query tool can execute arbitrary SQL, including DDL/DML, so most BigQuery operations are reachable indirectly. The main gaps are convenience features like listing datasets or managing query jobs, but agents can work around these with SQL and INFORMATION_SCHEMA queries.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
The BigQuery remote MCP server is a fully managed service that uses the Model Context Protocol to connect AI applications and LLMs to BigQuery data sources. It provides secure, standardized tools for AI agents to list datasets and tables, retrieve schemas, generate and execute SQL queries through natural language, and analyze data—enabling direct access to enterprise analytics data without requiring manual SQL coding.
Query BigQuery, Snowflake, Redshift & Azure Synapse with natural language
The all-in-one data stack for agents. Upload files, run SQL, evolve tables, and render charts.
Explore and query your data warehouse through Myriade's AI data analyst agent.
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/SnowLeopard-AI/bigquery-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
