I’ve been working on PolyDB for over a year. As I did more and more with agents I would ‘discover’ I needed yet another data store, and then another to efficiently manage the data requirements. My professional background is in infrastructure initially physical and later like the rest of the world cloud (aws). So as I tumbled down the rabbit hole I would add another service, which led me to the question: surely these models could be combined into a single database/data layer for an agent, letting it figure out how best to use the available tools.
From that came PolyDB, 16 data models/69 MCP tools/1 api key. I’ve been using PolyDB for my internal projects, agents, services, and platforms and I think it works pretty good for me. At this point it is NOT an enterprise database, although yeah I would like it to be. For example cross-model transactions aren't there yet. I’ve taken it as far as I can with my projects and although it's far from finished (is anything ever?), after lots and lots of testing I believe it works and have found/fixed as much as I can based on my use cases.
So who is PolyDB for? I’m thinking it’s a data store for an autonomous/near autonomous agent. An agent capable enough to ‘understand’ the available mcp tools and use them. Can I support other use cases? Certainly.
PolyDB doesn’t replace someone using niche vendor-specific functionality in their data layer of choice. I thought about the 80/20 rule in this case: maybe 80% of people are using 20% of the functionality and could PolyDB address that 20% across multiple data models, and then grow as use cases and gaps were identified.
At this point the PolyDB interface is MCP-driven; as it grows adding an SDK is on the roadmap. But it’s really driven by what people are looking for.
I have NO VC backing, as such there isn’t a free plan (because everything costs something). To be non-predatory, I do have a small plan that does not auto-renew. So you can test it out without fear of a recurring charge.
Like I said, I've been working on this for over a year, and it works for me, let me know where it breaks for you.
richardwii•50m ago
From that came PolyDB, 16 data models/69 MCP tools/1 api key. I’ve been using PolyDB for my internal projects, agents, services, and platforms and I think it works pretty good for me. At this point it is NOT an enterprise database, although yeah I would like it to be. For example cross-model transactions aren't there yet. I’ve taken it as far as I can with my projects and although it's far from finished (is anything ever?), after lots and lots of testing I believe it works and have found/fixed as much as I can based on my use cases.
So who is PolyDB for? I’m thinking it’s a data store for an autonomous/near autonomous agent. An agent capable enough to ‘understand’ the available mcp tools and use them. Can I support other use cases? Certainly.
PolyDB doesn’t replace someone using niche vendor-specific functionality in their data layer of choice. I thought about the 80/20 rule in this case: maybe 80% of people are using 20% of the functionality and could PolyDB address that 20% across multiple data models, and then grow as use cases and gaps were identified.
At this point the PolyDB interface is MCP-driven; as it grows adding an SDK is on the roadmap. But it’s really driven by what people are looking for.
I have NO VC backing, as such there isn’t a free plan (because everything costs something). To be non-predatory, I do have a small plan that does not auto-renew. So you can test it out without fear of a recurring charge.
Like I said, I've been working on this for over a year, and it works for me, let me know where it breaks for you.