VISTA Data MCP Server
Allows running read-only SQL queries (SELECT, SHOW, DESCRIBE, EXPLAIN) against ClickHouse databases for telemetry data.
Allows searching and retrieving documents from Elasticsearch indices for download analytics.
Click on "Deploy 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., "@VISTA Data MCP Server list all databases in ClickHouse"
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.
VISTA Data MCP Server
Read-only MCP server for querying ClickHouse (product telemetry) and Elasticsearch (download analytics). Built for Percona VISTA.
Install
curl -fsSL https://raw.githubusercontent.com/Percona-Lab/vista-data-mcp/main/install-vista-data-mcp | bashThe installer will:
Install
uvif neededAsk you to choose Remote (default) or Local mode
Configure Claude Desktop and Claude Code automatically
Install the VISTA plugin
Remote mode — connects to the shared Percona server over VPN. No credentials, no clone, no local install. Most users choose this.
Local mode — for users with their own ClickHouse or Elasticsearch credentials who want to run off-VPN (development, custom data sources).
Re-run the same command to update or change mode.
Related MCP server: db-connect-mcp
Tools
ClickHouse (telemetry)
Tool | Description |
| Run any read-only SQL (SELECT, SHOW, DESCRIBE, EXPLAIN) |
| Show all accessible databases |
| Show tables in a database |
| Show column names, types, and comments |
| Get sample rows from a table |
Elasticsearch (downloads)
Tool | Description |
| Run an Elasticsearch query (JSON DSL) |
| Show all indices with doc counts and sizes |
| Show field mapping (schema) for an index |
| Get sample documents from an index |
All results are returned as Markdown tables. Only read-only queries are allowed.
Manual Install (alternative)
If you prefer not to use the installer:
Claude Code (CLI)
claude mcp add vista-data \
-e CLICKHOUSE_HOST=your-ch-host \
-e CLICKHOUSE_PORT=8443 \
-e CLICKHOUSE_USER=default \
-e CLICKHOUSE_PASSWORD=your-ch-password \
-e CLICKHOUSE_DATABASE=default \
-e CLICKHOUSE_SECURE=true \
-e ES_HOST=your-es-host \
-e ES_PORT=9200 \
-e ES_USER=your-es-user \
-e ES_PASSWORD=your-es-password \
-e ES_SECURE=true \
-- uvx --from git+https://github.com/Percona-Lab/vista-data-mcp vista-data-mcpDevelopment
git clone https://github.com/Percona-Lab/vista-data-mcp.git
cd vista-data-mcp
cp .env.example .env
# Edit .env with your credentials
uv run mcp_server.pySafety
Read-only: ClickHouse allows only
SELECT,SHOW,DESCRIBE,EXPLAIN. Elasticsearch uses search API only (no indexing).Mutation blocking:
INSERT,UPDATE,DELETE,DROP,ALTER,CREATE,TRUNCATEare rejectedRow/hit limits: Results capped at 500 (configurable via
CLICKHOUSE_MAX_ROWS/ES_MAX_HITS)Query timeout: 30 seconds default (configurable via
CLICKHOUSE_QUERY_TIMEOUT/ES_QUERY_TIMEOUT)Identifier sanitization: Database and table names validated against injection
Environment Variables
ClickHouse
Variable | Required | Default | Description |
| Yes* | — | ClickHouse server hostname |
| No |
| HTTP(S) port |
| No |
| Username |
| No |
| Password |
| No |
| Default database |
| No |
| Use HTTPS |
| No |
| Timeout in seconds |
| No |
| Max rows per query |
Elasticsearch
Variable | Required | Default | Description |
| Yes* | — | Elasticsearch hostname |
| No |
| Port |
| No |
| Username (for basic auth) |
| No |
| Password |
| No |
| Use HTTPS |
| No |
| Verify SSL certificates |
| No |
| Timeout in seconds |
| No |
| Max hits per search |
*Only required if you want to use that data source. Tools for unconfigured sources return a helpful message.
Part of the Alpine Toolkit
License
Apache 2.0
Available Tools
9 toolsch_describe_tableA
Show the schema (columns, types, comments) of a ClickHouse table.
Args: table: Table name. database: Database name. If omitted, uses the default database.
Inputs: table (required), database. Returns: result
ch_list_databasesA
List all databases accessible in the ClickHouse instance.
Inputs: no named parameters. Returns: result
ch_list_tablesA
List all tables in a ClickHouse database.
Args: database: Database name. If omitted, lists tables in the default database.
Inputs: database. Returns: result
ch_sample_dataA
Get sample rows from a ClickHouse table (up to 100 rows).
Useful for understanding telemetry data structure before writing queries.
Args: table: Table name. database: Database name. If omitted, uses the default database. limit: Number of rows to return (1-100, default 10).
Inputs: table (required), limit, database. Returns: result
es_get_mappingA
Show the field mapping (schema) for an Elasticsearch index.
Args: index: The index name to inspect.
Inputs: index (required). Returns: result
es_list_indicesA
List all Elasticsearch indices with document counts and sizes.
Use this to discover available download/package data indices.
Inputs: no named parameters. Returns: result
es_sample_dataA
Get sample documents from an Elasticsearch index (up to 100).
Useful for understanding download data structure before writing queries.
Args: index: The index name. size: Number of documents to return (1-100, default 10).
Inputs: index (required), size. Returns: result
query_clickhouseA
Run a read-only SQL query against ClickHouse (telemetry data).
Only SELECT, SHOW, DESCRIBE, and EXPLAIN statements are allowed. Results are capped at 500 rows by default.
Use this for product telemetry: active instances, version distribution, storage engines, deployment types, CPU architecture, cluster metrics.
Args: sql: A read-only SQL statement.
Examples: - SELECT count() FROM telemetry WHERE product = 'MySQL' - SELECT version, count() as n FROM telemetry GROUP BY version ORDER BY n DESC - SHOW TABLES
Inputs: sql (required). Returns: result
search_elasticsearchA
Run an Elasticsearch query (JSON DSL) against download/package data.
Use this for product download data: downloads by product, package type, OS, components, growth rates, EOL packages, Pro-builds.
Args: index: The Elasticsearch index to search. query_body: A JSON string containing the Elasticsearch query DSL. Supports match, term, range, bool, aggregations, etc. size: Max documents to return (default 500).
Examples: - {"query": {"match": {"product": "postgresql"}}, "size": 20} - {"size": 0, "aggs": {"by_package": {"terms": {"field": "package_type"}}}} - {"query": {"range": {"date": {"gte": "2025-01-01"}}}}
Inputs: index (required), query_body (required), size. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 9 tools
Each tool is clearly scoped to either ClickHouse or Elasticsearch via distinct prefixes (ch_/es_) and a distinct action (list, describe, sample, query). Within each backend, there is no overlap between listing resources, inspecting schema, sampling data, and executing queries.
The ch_ and es_ prefixes provide a consistent pattern for most tools, but the generic query tools (query_clickhouse, search_elasticsearch) break the pattern by placing the verb first and omitting the prefix. Additionally, 'describe' vs 'get' are used inconsistently across backends, but the overall system remains readable.
Nine tools is a well-scoped set for a dual-backend data exploration server. Each tool serves a distinct purpose, covering discovery, schema inspection, sampling, and querying for both ClickHouse and Elasticsearch without excess.
The server provides a complete read-only workflow for both data sources: listing available resources (databases/indices/tables), inspecting schemas/mappings, previewing data, and running arbitrary queries. No critical operations are missing for its stated purpose.
Related MCP Connectors
MCP server for querying and analyzing data from ad platforms, analytics tools, and spreadsheets
Read-only Google Analytics 4 MCP server for AI assistants.
Read-only MCP server for The Quiet Protocol's engines, benchmarks, proof, and business data.
Read-only MCP server for AIStatusDashboard status, incidents, metrics, and fallback recommendations.
Related MCP Servers
- AlicenseBqualityDmaintenanceA read-only MCP server for OpenObserve Community Edition that works over the REST API. Provides tools for searching logs, traces, stream schemas, and dashboards - no Enterprise license required.873 PyPI16GPL 3.0
- AlicenseAqualityBmaintenanceA read-only MCP server for exploratory data analysis across PostgreSQL, MySQL, and ClickHouse databases, providing safe, read-only access with comprehensive analysis capabilities.1084 PyPI6MIT
- AlicenseNot gradedqualityCmaintenanceA read-only MCP server for querying telemetry data from configurable backends. Provides tools to list sources, describe schemas, run bounded queries, and compute aggregates.MIT
- AlicenseAqualityCmaintenanceReadonly PostgreSQL MCP server with SQL guardrails for analytical queries and schema introspection.221 npmMIT