# GoNavi: Lightweight Multi-Database Client with Native MCP and AI Support

> GoNavi is a cross-platform database client built with Wails (Go) and React that ships 20-26 MB installers by using the operating system's native WebView instead of Electron. It covers SQL, Redis, Kafka, vector, time-series, and several domestic Chinese databases, and includes a built-in MCP server for connecting AI agents to database schema context.

**Syngnat/GoNavi** — High-performance multi-data-source database client — ~30MB, AI & MCP ready, zero Electron bloat. | 高性能多数据源数据库客户端：约 30MB，AI 与 MCP 就绪，告别 Electron 膨胀。

- Repository: https://github.com/Syngnat/GoNavi
- Website: https://gonavi.org
- Stars: 1,906 · Forks: 207
- Language: TypeScript
- License: Apache-2.0
- Published: 2026-09-10 · Updated: 2026-09-10 · Language: en
- Canonical page: https://hysenlabs.com/projects/syngnat-gonavi

## What GoNavi Addresses and Who It Targets

Desktop database clients built on Electron ship large installers because they bundle a full Chromium browser engine with every release. A typical Electron database client installer runs into the hundreds of megabytes. This is an acknowledged trade-off in the Electron model: the runtime is identical everywhere, but users pay the download cost.

GoNavi takes a different path. It uses Wails, a Go framework that builds desktop applications using the operating system's native WebView (WebKit on macOS and Linux, WebView2 on Windows) rather than a bundled Chromium. The frontend is React. The result is an installer in the 20-26 MB class for Windows, macOS, and Linux release assets.

The trade-off is explicit in the README: the installer size is small, but the idle RSS on a Linux build (using the WebKit build, empty workbench, no database connections) was measured at approximately 429 MB for the main process and approximately 765 MB including WebKit helper processes. The README labels this as a single measurement from a Linux cloud run of the v0.9.8 WebKit41 build and cautions against treating it as a benchmark or comparing it to unverified marketing claims.

GoNavi targets developers and database engineers who need a single client across SQL, cache, vector, message queue, search, time-series, and domestic databases, and who are building or maintaining AI-assisted workflows where structured database context must be available to coding agents.

## Architecture: Go Backend, React Frontend, Native WebView

The application is a Wails v2 project. The Go backend handles all database driver communication, and the React frontend provides the query editor, data grid, and connection management UI. The go.mod file lists an extensive set of database drivers: ClickHouse, MySQL (go-sql-driver/mysql), MongoDB (mongo-driver), PostgreSQL (lib/pq), Microsoft SQL Server (go-mssqldb), Oracle (go-ora), Redis, Kafka (segmentio/kafka-go), TDengine (driver-go), IoTDB, Pulsar, RocketMQ, MQTT, Elasticsearch, and several Chinese domestic databases including OceanBase, DM (Dameng), GaussDB, KingBase, and Trino.

The workbench layout separates three primary areas: a connections panel for managing SSH proxies and driver configuration, a Monaco-based SQL editor with schema context and slash commands for AI assistance, and a virtualised data grid for viewing and editing results with transaction submit and rollback support.

MCP (Model Context Protocol) is built into the application as a first-class server, not a plugin. The go.mod includes the official modelcontextprotocol/go-sdk dependency. Docker Compose files for deploying the MCP server in standalone mode are included in the repository: Dockerfile.mcp-server and docker-compose.mcp-server.yml.

## Getting Started: Downloading and Connecting

GoNavi distributes pre-built installers for Windows, macOS, and Linux through the GitHub Releases page and the project website at gonavi.org. The README describes these as approximately 20-26 MB class installers. Version 1.0.0 was released on 2026-09-24, with a dev build following on 2026-09-28.

For users wanting to understand the workbench layout before downloading, the README includes this ASCII overview of the three-panel structure:

```text
┌──────────────────────────────────────────────────────────────────────┐
│  GoNavi Workbench                                                    │
│  ┌─────────────┐  ┌──────────────────┐  ┌─────────────────────────┐  │
│  │ Connections │  │ Monaco SQL + AI  │  │ Virtualized DataGrid    │  │
│  │ SSH / Proxy │  │ Schema context   │  │ Batch edit · Export     │  │
│  │ Drivers     │  │ Slash commands   │  │ Txn submit / rollback   │  │
│  └─────────────┘  └────────┬─────────┘  └─────────────────────────┘  │
```

Connections in the left panel can route through SSH proxies, which the README lists as part of the connection configuration. The AI panel in the Monaco editor uses the connected database's schema as context when generating or explaining queries, which keeps schema information local to the GoNavi instance rather than transmitting passwords or credentials to external services.

## MCP Integration for AI Coding Agents

GoNavi includes a built-in MCP server that allows AI coding agents to query database schema and run read operations against connected databases. The MCP server can be run as part of the desktop application or deployed separately using the dedicated Docker Compose configuration.

The project's mcpvault.io health badge in the README indicates the MCP server is registered and monitored on the MCPVault registry. The go.mod dependency on modelcontextprotocol/go-sdk confirms the implementation follows the standard Go SDK for the protocol.

This matters for teams using AI coding assistants that support MCP: the agent can request schema information, table structures, and column definitions from GoNavi without the human operator manually copying schema output. The README states that the GUI retains ownership of schema modifications, edits, and EXPLAIN operations, with the AI assistant handling drafting. This is a design boundary: MCP-connected agents get read and draft access, not write access through the protocol.

## Database Coverage and Domestic Database Support

The go.mod dependency list reveals the full scope of driver support. Beyond the standard relational databases (MySQL, PostgreSQL, SQLite, SQL Server, Oracle), GoNavi includes drivers for Milvus vector database, ClickHouse columnar store, Kafka and Pulsar message queues, MQTT for IoT workloads, Elasticsearch for search, IoTDB and TDengine for time-series data, and Redis for cache and key-value.

The domestic Chinese database coverage is notable for a tool at this stage: DM (Dameng, a PL/SQL-compatible database used in Chinese government systems), GaussDB (Huawei's distributed database), KingBase, OceanBase, and Trino. This breadth makes GoNavi relevant for teams working on systems that span international and domestic Chinese database deployments.

DuckDB is also in the go.mod via duckdb/duckdb-go, which enables in-process analytical queries directly in GoNavi without requiring a separate database server.

## Limitations and Things to Verify

GoNavi's default branch is dev, not main. Version 1.0.0 was tagged on 2026-09-24, making this a newly released product. The early release status means the feature set and stability characteristics are less proven than tools with years of production use behind them.

The installer size advantage requires the operating system's native WebView to be present and functional. On Windows, WebView2 must be installed (it ships by default with Windows 11 and most Windows 10 systems, but not all). On Linux, the WebKit build requires a running display server; the RAM figures in the README were measured on a remote Linux cloud VM with a remote display, which may differ from desktop deployments.

The Apache-2.0 licence requires attribution in derivative works and documentation of any modifications made. This is a permissive licence but has more requirements than MIT. The NOTICE file in the repository contains the required attribution text.

## Sponsorship Model and Maintenance

GoNavi is sponsored by APISmart and HuaLongAI, both of which are AI API proxy services. The README includes a detailed sponsor table explaining what each provides and why it fits GoNavi's AI integration use case. This sponsorship model means the project's continued development is partially supported by commercial relationships rather than pure community contributions.

The last push to the repository was on 2026-09-15. The v1.0.0 release was published on 2026-09-24. The project is in early active development and has been receiving frequent updates. The default branch being dev rather than main is a conventional indicator that the team considers the main line of development to still be in flux.

Contributions are documented in CONTRIBUTING.md and its Chinese counterpart CONTRIBUTING.zh-CN.md. The repository includes AGENTS.md and CLAUDE.md files indicating the maintainers use AI coding agents in their development workflow and have codified rules for them.

## Conclusion

Development teams who work across multiple database types and want a single client without the memory and installer overhead of Electron-based tools will find GoNavi a technically substantive option. Its MCP integration for AI coding agents is a first-class feature, not an add-on. Users considering GoNavi should note that the default branch is dev, that version 1.0.0 was released on 2026-09-24, and that the project is newly released. The RAM figures in the README (idle Linux RSS of approximately 429 MB for the main process, 765 MB with WebKit helpers) are the honest cost of the system WebView approach on Linux, not the installer size. Verify your target platform's WebView behaviour before comparing these numbers to Electron.

## FAQ

### How does GoNavi compare to Electron-based database clients in installer size?

GoNavi ships approximately 20-26 MB class installers by using the operating system's native WebView (WebKit on macOS and Linux, WebView2 on Windows) instead of bundling Chromium. However, the idle RAM usage on Linux (approximately 429 MB for the main process, 765 MB with WebKit helpers) should be evaluated separately from the installer size, as the README cautions.

### What databases does GoNavi support?

GoNavi's go.mod includes drivers for MySQL, PostgreSQL, SQL Server, Oracle, MongoDB, SQLite, Redis, Kafka, Pulsar, RocketMQ, MQTT, ClickHouse, Milvus, Elasticsearch, TDengine, IoTDB, DuckDB, Trino, OceanBase, and several domestic Chinese databases including DM (Dameng), GaussDB, and KingBase.

### How does GoNavi's MCP integration work?

GoNavi includes a built-in MCP server that exposes database schema and read access to AI agents that support the Model Context Protocol. The MCP server can run embedded in the desktop app or be deployed as a standalone Docker container. The README states that schema modifications and EXPLAIN operations remain under the GUI's control, not the MCP-connected agent.

## Sources

- [License: Apache-2.0](https://github.com/Syngnat/GoNavi/blob/dev/LICENSE)
- [Project website](https://gonavi.org)
- [README](https://github.com/Syngnat/GoNavi/blob/dev/README.md)
- [Releases](https://github.com/Syngnat/GoNavi/releases)
- [Syngnat/GoNavi on GitHub](https://github.com/Syngnat/GoNavi)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/syngnat-gonavi