👉👉👉 Please share your use case to shape DBHub roadmap

[tianzhou]

Hey there! 👋

I'd love to hear how you're using DBHub (or planning to use it). Your feedback will directly influence our roadmap priorities.

Please share:

• What's your primary use case? (e.g., local development, data exploration, AI/LLM integration, automation scripts)
• What's working well?
• What's missing or could be improved?

[DmitriyAlergant]

@tianzhou:

Primary use case: local development, data exploration, production support/troubleshooting

What's working well: this MCP server only having two tools (execute_sql and search_object) and minimally loading the context window. Please keep it this way. Prior to that I tried and thrown away two or three other postgres MCP tools which were insanely overblown. Minimal context load is crucial for MCP. execute_sql is more then enough for most things, and search_object is a nice addition.

What's missing or could be improved:

• improve README.md documentation with precise instructions on mcp installation to different clients (preferably over stdio where applicable) for dumb frictionless copy&paste (bring your own database url). See example on https://github.com/upstash/context7 You are both MIT licensed I think you can just borrow the content.
• discoverability. Not sure how to achieve that. Your tool was not easy to find. It is not coming up on most searches by "postgres mcp", neither in google nor on awesome mcp servers (https://mcpservers.org/) etc; You are there (https://mcpservers.org/servers/bytebase/dbhub) if somehoe searches by "dbhub" but not findable by postgres.... I only found you through word of mouth.

Universal database MCP server supporting mainstream databases - well, your motto might be that this MCP is 'minimalistic' for postgres mssql mysql sqlite; "Minimalistic" is a differentiator, "universal" is not...

Thanks!

[tianzhou]

@DmitriyAlergant the minimalistic suggestion is spot on. Taken verbatim.

[erbanku]

Our client's company has very old Oracle databases, some of which have nearly 100 columns in the worst case. Cannot find any useful AI-integrated (especially MCP) tools yet.
It could be wonderful if dbhub can support Oracle DB at least in the first stage. And this could be very useful to dbhub itself, as there are few or no tools that support that DB.

I also posted this request here: #223

I hope it will be considered!

[marc-odp]

Since dbhub is multi-databases, LLM get confused sometimes about which database / table to use in a certain context. This can be solved by giving information to the AI but maybe it would be nice to be able to add description directly in the MCP ?

[tianzhou]

@marc-odp have you tried specify the description for the tool https://dbhub.ai/config/toml#description

[marc-odp]

I was not talking about 'Tools' but about 'Sources'. Is it also possible to add a description to a Source ?
A 'Source' can contain many databases and it is not possible for the LLM to select the right one considering the context only.

@marc-odp have you tried specify the description for the tool https://dbhub.ai/config/toml#description

[tianzhou]

@marc-odp added in #232

[cyreb7]

I use it for local development inside a Docker container with Claude Code. Primarily used by Claude for querying to check the DB schema, and sometimes for verifying SQL before adding it to code. But occasionally I ask it to debug specific data issues or get statistics about what's in the DB.

It still can't replace DBeabver for two use cases:

1. Multiple connection profiles to directly connect to the dev/prod DB (not sure that I'd want to give AI access for this)
2. Occasional write access (I never need this for normal development and am happy copying queries and running them myself)

But it works great for normal development. Love the low context use and being able to give Claude unconditional tool usage because it's read-only.

[tianzhou]

@cyreb7 while DBHub couldn't replace DBeaver for that 2 use cases, if you happen to use Postgres, you can check out my another project pgconsole: https://github.com/pgplex/pgconsole

[jankogasic]

Hi. Is there a way to support multi-tenancy? I want to have one instance od dbhub which will serve multiple users where each one has its own set of permissions over my database.

[tianzhou]

@jankogasic Currently DBHub does not have built-in multi-tenancy or per-user permission support. Each instance runs with a single set of database credentials and tool-level settings (like readonly, max_rows).

Here's what exists today and what the gap is:

What you can do now:

• Run separate DBHub instances per user, each with its own TOML config pointing to the same database but using
  different DB credentials (with DB-level permissions)
• Use readonly = true and max_rows in TOML [[tools]] sections — but these apply globally, not per-user

Why per-user multi-tenancy is hard with the current design:

• The MCP protocol (stdio and streamable HTTP) doesn't have a built-in authentication/identity layer
• DBHub's HTTP transport runs stateless — there's no session or user context passed with requests
• The connector manager maps source IDs to connections, not users to permissions

[atlantis]

Hey thanks for a great tool! I'd really like the ability to validate (but not execute) queries... I'm using mysql and I could ask the model to prepend EXPLAIN to my query but it would be nicer/safer if some sort of validation or explain functionality were built in.

If it were explain, it would also be useful because the model could diagnose performance problems with queries!

[xkaaaaaa]

Is there a roadmap for Redis support?