| FMRS | 55 / 100 · C | 79 / 100 · B | 75 / 100 · B |
|---|
| Reliability | 10 / 20 | 12 / 20 | 12 / 20 |
|---|
| Security and permissions | 8 / 20 | 18 / 20 | 16 / 20 |
|---|
| Maintenance | 12 / 20 | 17 / 20 | 14 / 20 |
|---|
| Documentation | 13 / 20 | 17 / 20 | 18 / 20 |
|---|
| Setup experience | 12 / 20 | 15 / 20 | 15 / 20 |
|---|
| Best for | - Data analysts and engineers who want to query and explore Apache Druid data via AI assistants like Claude, ChatGPT, or Gemini using natural language.
- Teams looking to automate or streamline Druid cluster management tasks such as monitoring tasks, compaction, retention rules, and ingestion.
- Developers integrating MCP clients via supported transports (STDIO, Streamable HTTP) looking for a standard 'out-of-the-box' server to interact with Druid.
- Users who want a conversational data interface in a local desktop application such as Ypipe.
| - Teams already using ClickHouse who want AI assistants to access data directly.
- Scenarios requiring fast, read-only data queries and schema exploration.
| - Drug discovery and computational chemistry researchers
- Bioinformatics teams needing ChEMBL data through AI assistants
- Developers comfortable with Node.js or Docker
|
|---|
| Not for | - Not for latency-sensitive real-time data operations when MCP session overhead is unacceptable.
- Not for headless, non-interactive scenarios where MCP protocol abstraction is unnecessary.
- Not for users who require deep, custom Druid API capabilities or complex workflows that are not exposed through the MCP.
- Not for those looking for a server to interoperate with cloud-hosted services such as Imply Polaris.
| - Scenarios requiring write access to the database without explicit opt-in.
- Production environments with stringent security requirements that avoid default permission settings.
| - Production settings requiring official ChEMBL support (this is unofficial)
- Users without a Node.js or Docker environment
- General database tasks unrelated to ChEMBL
|
|---|
| Required permissions | - Must be provided with URLs for Druid Router (required) and Coordinator (recommended) via environment variables.
- Optionally provide Druid username and password for authentication via HTTP Basic.
- For HTTP transports, OAuth2 is enabled by default and requires clients to present a valid Bearer token.
- Varying permissions are required for admin operations depending on which Spring profiles are active (`ops`, `permissions`, `health`).
- Read access to Druid system tables/catalog and data access depending on the scope of queries.
| - Requires read-only access to ClickHouse database (default).
- Optional: write access via CLICKHOUSE_ALLOW_WRITE_ACCESS.
- Optional: destructive operations via CLICKHOUSE_ALLOW_DROP.
| - Outbound network access to https://www.ebi.ac.uk/chembl/api/data
- Read-only access to public API data; no authentication credentials required
|
|---|
| Risks and side effects | - Risk of unintentional data deletion or modification when the `ops` profile is active and tokens are compromised, as `manageDatasourceOrSegment` allows permanently dropping datasources.
- Using SSL with disabled certificate verification (`DRUID_SSL_SKIP_VERIFICATION=true`) can expose connections to man-in-the-middle attacks.
- Exposing sensitive cluster metadata and query results to AI assistants may lead to data leakage.
- The `permissions` tools can modify users, roles, and authorization policies when proper permissions are held.
- Because tools are activated on-demand based on Spring profiles, there is a risk of misconfiguration if the required profiles are not activated correctly.
| - 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.
| - Unofficial project with license marked NOASSERTION (README claims MIT); verify licensing yourself
- Data accuracy depends on the ChEMBL REST API and the server's parsing may contain defects
- 30-second request timeouts may constrain large batch queries
- README inconsistency on tool count (22 vs 27) raises documentation reliability concerns
|
|---|
| Supported clients | Claude Desktop, Generic MCP Client | Claude Desktop | Claude Desktop |
|---|
| Tools | 29 | 4 | 27 |
|---|