Wanaku 0.2.0 brings live SQL querying to LLMs via MCP

Otavio Rodolfo Piske's post addresses a core limitation of AI assistants: their knowledge is frozen at training time, making them unreliable for live operational data. While RAG (Retrieval-Augmented Generation) with vector databases works well for static text documents and wikis, it is poorly suited to data that changes too fast for vector syncs — such as inventory levels, customer orders, or real-time application metrics. The `sql-tool` service template in Wanaku's upcoming 0.2.0 release targets exactly this gap by letting LLMs issue on-demand SQL queries against live relational databases through the Model Context Protocol (MCP).

Wanaku is described as an open-source MCP router and capability management platform that acts as a bridge between AI clients — such as Claude Desktop or Continue.dev — and backend services.

Wanaku is described as an open-source MCP router and capability management platform that acts as a bridge between AI clients — such as Claude Desktop or Continue.dev — and backend services. Version 0.2.0 introduces the concept of service templates: pre-packaged capability patterns instantiated via a single CLI command. The `sql-tool` template bundles an Apache Camel route, an MCP tool definition for dynamic client discovery, pre-packaged runtime dependencies (including JDBC drivers, SQL components, and JSON marshalling), and support for Camel Simple expressions to inject AI inputs into queries at runtime. The post walks through a PostgreSQL example using Podman, where the `wanaku service template instantiate` command is given a query like `SELECT name, price FROM products WHERE price < ${body} ORDER BY price`; the `${body}` placeholder is replaced at runtime with the value the AI passes in, making the tool interactive rather than static. Service registration can be verified with `wanaku service catalog list` and data store connectivity with `wanaku data-store list --plain`.

Key facts

1. 01Wanaku's `sql-tool` service template is coming in version 0.2.0, with PR #1358 pending merge.
2. 02The template exposes live relational database queries as executable MCP tools discoverable by any MCP-compatible AI client.
3. 03It uses Apache Camel routes with Camel Simple expressions (e.g., `${body}`) to inject AI inputs into dynamic SQL queries at runtime.
4. 04Pre-packaged dependencies include JDBC drivers, SQL components, and JSON marshalling.
5. 05The approach targets live operational data (inventory, orders, metrics) that changes too fast for RAG-style vector pre-indexing.
6. 06The post demonstrates a PostgreSQL example run via Podman, using the `wanaku service template instantiate` CLI command.
7. 07MCP-compatible clients such as Claude Desktop and Continue.dev can automatically discover and call the registered tool.

Topics