| FMRS | 62 / 100 · C | 79 / 100 · B | 74 / 100 · B |
|---|
| Reliability | 8 / 20 | 12 / 20 | 13 / 20 |
|---|
| Security and permissions | 16 / 20 | 18 / 20 | 13 / 20 |
|---|
| Maintenance | 10 / 20 | 17 / 20 | 17 / 20 |
|---|
| Documentation | 15 / 20 | 17 / 20 | 16 / 20 |
|---|
| Setup experience | 13 / 20 | 15 / 20 | 15 / 20 |
|---|
| Best for | - Developers and analysts who want to query BigQuery quickly via AI assistants
- Organizations that need to prevent AI agents from modifying data
- Environments with sensitive data that require field-level access restrictions
| - Teams already using ClickHouse who want AI assistants to access data directly.
- Scenarios requiring fast, read-only data queries and schema exploration.
| - Teams that want precise control over which database operations an AI can perform, rather than open arbitrary SQL execution
- Scenarios needing a unified MCP setup across multiple database engines
|
|---|
| Not for | - Scenarios requiring write, update, or delete operations
- Users who need unrestricted access to all columns
- Strict security environments that cannot tolerate data leaving the network (since results are sent to LLM providers)
| - Scenarios requiring write access to the database without explicit opt-in.
- Production environments with stringent security requirements that avoid default permission settings.
| - Lightweight cases that just want to run a few ad-hoc SQL queries without maintaining a tools.yaml config (a simpler single-database MCP may be a better fit)
|
|---|
| Required permissions | - BigQuery read permissions (e.g., roles/bigquery.user or roles/bigquery.dataViewer)
- For service account auth, access to the service account key file
- For advanced features, permission to list datasets and tables, and run dry-run queries
| - Requires read-only access to ClickHouse database (default).
- Optional: write access via CLICKHOUSE_ALLOW_WRITE_ACCESS.
- Optional: destructive operations via CLICKHOUSE_ALLOW_DROP.
| - Database credentials (username/password/connection string) are supplied via env vars or config
- A tool's actual permission is whatever SQL statement is defined in tools.yaml — designed for least privilege, but misconfiguration can still over-expose access
|
|---|
| Risks and side effects | - Data exfiltration: query results are sent to LLM providers (Anthropic, OpenAI) and may leave your network
- Despite field restrictions, AI agents may attempt to infer data through crafted queries
- Misconfiguration may weaken restrictions, e.g., incorrectly setting preventedFields or maximum bytes billed
| - If write access is enabled, AI might make unintended modifications.
- If DROP access is enabled, data deletion could occur accidentally.
- Credentials may be exposed via environment variables.
| - If tools.yaml defines SQL statements that allow unconstrained writes or deletes, the AI could accidentally modify data
- The prebuilt toolsets (--prebuilt) favor convenience and may expose broader query capability than a specific business actually needs — use a custom tools.yaml in production
|
|---|
| Supported clients | Claude Desktop, Claude Code | Claude Desktop | Claude Code, Gemini CLI, Zed, Antigravity |
|---|
| Tools | 0 | 4 | 0 |
|---|