Convo is the fastest way to log, debug, and personalize AI conversations. Capture every message, extract long-term memory, and build smarter LLM agents - with one drop-in SDK.
We built Convo SDK after months of wrangling LangGraph checkpointers and database infra just to get persistent memory working.
Every time we added memory to an agent, we ended up knee-deep in connection pools, schema migrations, and random production crashes 😵💫
So we made something simple:
One line to replace any LangGraph checkpointer.
No Postgres. No Mongo. No ops.
Just:
checkpointer: convo.checkpointer()
It’s TypeScript-first, handles multi-user threads out of the box, and gives you time-travel debugging + bulletproof persistence, all without spinning up a DB.
We’d love feedback from other agent builders, LangGraph users, or anyone tired of “database hell.”
Excited to hear what you think!
_________________
🔧 Want to try it in your own stack?
We put together a quick-start cookbook with real-world recipes to get going fast:
@sunnyjoshi one line to replace LangGraph checkpointer? Brilliant - I’ve been struggling with similar memory issues in my Claude Code sessions. Does this work directly with Anthropic’s API or would I need a wrapper? Database hell struggle is real 😆
The drop-in SDK and one-liner setup is beautifully dev-friendly :) Congratulations @sunnyjoshi and team!
I like that it’s TypeScript-first and supports multi-user threads without extra config. That alone is a reason for me to give it a shot.
Giving LLM apps memory and observability is key to making them smarter and more reliable — Convo looks like a must-have for any AI dev stack. Congrats on the launch 🚀
Logging every convo AND adding long-term memory? That’s huge for making LLM apps feel way smarter, tbh. You folks nailed it with this one!
We're also doing a virtual event covering memory and persistent graph state using checkpointers, please feel free to join - https://lu.ma/fl29ul0l
