Memory Store

i’ve been using beeper mcp with memory store

i’m shubham. i work across three completely different worlds - infrastructure for physical agents, a small coffee business, and software consulting gigs. i’m just out of college i need to make money so i do things.

my work spectrum is weird. one end is enterprise clients with proper contracts, other end is people messaging me on whatsapp coordinating coffee things. middle is exploratory - trying things, reading papers, connecting ideas.

most conversations happen on whatsapp, that’s just india. slack rarely, some linkedin and with discord with 1-2 people. mostly whatsapp though.

maybe 15-20 people i’m actively talking to. each conversation is a different universe.

how i was managing things (i wasn’t)

i’d maintain projects in chatgpt. or maybe i’d use claude somewhere. or free gemini but with a different email so i don’t even know where that is now different accounts everywhere mostly.

i use granola for meetings. i’d dump transcriptions there i used notebooklm.

i genuinely didn’t know where things were. which conversation happened where. which context was in which tool.

ai is only as useful as what you feed it. but i’m not supposed to be maintaining things. it should just get it. gpt is smart. granola is smart - actually makes my life easier as a product. but i was still the one connecting everything manually.

that doesn’t work.

things are changing. how we work should change with it.

what actually changed

i came across memory store a few months back. it’s an mcp server - that’s important. means it’s agnostic of where you’re working.

they call their approach “active memory” - creates connections and understands patterns on its own. not just storage. actual understanding.

i was already using beeper. everything consolidated there - whatsapp, slack, telegram. found out there’s a beeper mcp.

connected memory store to it mid-october.

now memory store sees my conversations. not just in beeper - everywhere i’ve added it. zed when i’m coding. vs code. claude desktop. chatgpt. raycast.

same memory layer. different tools.

replying to texts with context

here’s something that’s been useful - replying to texts through beeper mcp with memory store.

someone messages me about a project. memory store already knows what we talked about last week, what the context is, what matters to them. i can draft a response that actually makes sense without scrolling back through chat history.

especially useful when you’re juggling multiple conversations. i can be talking to someone about coffee sourcing, another person about infrastructure work, and memory store keeps the contexts separate but accessible.

what happens now

i don’t maintain anything. that’s the point.

after a meeting, granola has the transcription. memory store connects it to previous conversations, to other projects, to patterns i haven’t noticed.

someone messages me on whatsapp about coffee sourcing. memory store knows we talked about this before. knows the context. knows what matters to them.

i’m working on a client project in zed. memory store pulls relevant context from a conversation i had three weeks ago in a different tool.

it just works. not because i’m feeding it perfectly structured data. because it’s making connections autonomously.

here’s what i mean by it grew on me - i started with beeper. now it’s in zed, vs code, claude desktop, chatgpt, raycast. everywhere i work. because once you have memory that follows you across tools, you don’t want to work without it.

why this matters

i work across enterprise clients, whatsapp conversations about coffee, and exploratory technical work. three completely different contexts. different people. different problems.

i can’t maintain three separate systems. i can’t remember which detail lives where.

memory store as an mcp server means i don’t have to. it’s there wherever i’m working. beeper sees my conversations. my editors see my code. my ai tools see my thinking. memory store connects all of it.

for developers using beeper - this is actually amazing. you’re already context switching between code and conversations constantly. having memory that spans both is not a nice-to-have. it’s how we should be working.

what i have now

i don’t know where specific things are anymore. i don’t need to. memory store knows.

i can talk to 15-20 people across completely different industries and not lose context. i can switch between enterprise work, coffee business, and exploratory projects without maintaining separate systems.

it’s not about perfect organization. it’s about having memory that works how we actually work - scattered, jumping between tools, making connections we don’t consciously track.

ai should be smart enough to just get it. this setup does.

---

memory store is an mcp server in private beta. if you’re a developer using beeper and want to try this out, join the waitlist at memory.store. you can also dm me and i can probably get you access.

[i’ve used the memory store in writing this]