Show HN: Bind.ly – Persistent memory for AI across tools(bind.ly)
bind.ly
Show HN: Bind.ly – Persistent memory for AI across tools
https://bind.ly
4 comments
This tool intentionally has no usage analytics.
Feedback happens via a pinned GitHub Discussion instead:
paste a local telemetry summary, nothing else.
Thanks for the comment, veeduzyl — honestly just happy to see the first reply!
I might be missing something, but I’m exploring Bindly more as a shared, cloud-based memory layer that multiple chat agents can reference, rather than a single tool with its own analytics.
I’m still very early and figuring this out as I go.
By the way, it looks like you’re experimenting with similar constraints around feedback and telemetry — if your project is public, I’d genuinely be curious to check it out!
I might be missing something, but I’m exploring Bindly more as a shared, cloud-based memory layer that multiple chat agents can reference, rather than a single tool with its own analytics.
I’m still very early and figuring this out as I go.
By the way, it looks like you’re experimenting with similar constraints around feedback and telemetry — if your project is public, I’d genuinely be curious to check it out!
Thanks for taking the time to read this.
I’d really appreciate any kind of feedback — questions, skepticism, alternative approaches, or even “this wouldn’t work for me” takes. All of that is useful signal for me.
Happy to clarify anything or go deeper where helpful!
I’d really appreciate any kind of feedback — questions, skepticism, alternative approaches, or even “this wouldn’t work for me” takes. All of that is useful signal for me.
Happy to clarify anything or go deeper where helpful!
I’m a product designer by background, not a traditional software engineer.
Over the last year, tools like Claude and ChatGPT completely changed how I work. I started with small internal tools, and now I’m “vibe coding” multiple highly personalized apps.
As I iterated between Claude Code (implementation) and ChatGPT (ideation / thinking), I kept running into the same problem.
To think clearly, I had to repeatedly re-explain:
- what the code currently does,
- what changed recently,
- and why certain decisions were made.
That re-summarization step became a real bottleneck.
So I built Bindly (bind.ly).
Bindly is a persistent knowledge layer that sits outside any single AI tool.
The key idea is simple:
AI doesn’t remember. It re-reads shared context from the same place every time.
The concrete workflow I use today:
1) Claude Code → Bindly: After coding or refactoring, I ask Claude Code to summarize what changed and why, and store that context in Bindly.
2) ChatGPT → Ideation using Bindly: I then switch to ChatGPT and ideate based on that stored context — architecture, tradeoffs, next steps. Those ideation results are saved back into Bindly.
3) Claude Code → Reuse the ideation: Finally, I bring those ideation results back into Claude Code to continue implementation.
Bindly becomes the shared memory that closes this loop, without constantly restating everything.
To reduce both cognitive load and token usage, Bindly applies lightweight diffs (inspired by Git) and progressive search, so AIs only re-read what actually changed or what’s relevant right now.
In short:
- Bindly doesn’t try to replace AI thinking. It stores what exists, what was decided, and why - so any AI can continue from the same point.
Personally, this workflow already saves me a lot of cognitive overhead. But I’m unsure whether this is just a personal productivity hack or something others would actually pay for. I’m curious whether others who bounce between multiple AI tools run into the same problem.
Infrastructure uncertainty (and my move toward Cloudflare)
I initially built the MVP on Fly.io. It worked, and I don’t think Fly.io is a bad platform. But as Bindly grew, I became uncomfortable with how opaque things felt — volumes, persistence, failure modes.
Bindly is meant to be a knowledge layer. If data is lost, silently corrupted, or hard to reason about, that completely breaks the trust model.
For a tool whose purpose is “don’t lose context,” operational uncertainty felt like a product risk, not just a technical one.
I realized I was spending more time worrying about infrastructure behavior than thinking about the product itself.
So I started moving parts of the system onto Cloudflare, aiming to reduce operational uncertainty and keep the infrastructure as boring and predictable as possible.
So far, it feels simpler and more transparent. I’m still unsure whether going fully Cloudflare-first is the right long-term decision, but reducing cognitive overhead at the infrastructure level has already made a noticeable difference.
What I’d really appreciate feedback on:
1) Does a shared “knowledge / memory layer” like this feel useful beyond one person?
2) As a non-engineer, does moving toward a Cloudflare-first setup seem like a reasonable long-term direction?
Some extra context (optional):
I joined HN back in 2018 to apply to YC and interviewed in Nov 2018 — didn’t make it in the final round. This is my first time posting here in years.
Current ways I personally access Bindly:
- via Claude (Web)
- via ChatGPT APP (Web)
- via MCP (IDE)
If you have time to share thoughts - even critical ones — I’d be very grateful.
Happy New Year, and thanks for reading!