Peppermint

we put my coworker's AI twin on our website. you can chat with it right now. it knows what he's working on, what he said in standup, what's in his linear queue. it's 50% impressive and 50% deeply weird. peppermint.com/landing-page — go talk to liam.

Peppermint for Windows is here. It sees your screen, hears your meetings, and remembers your files — so it can draft emails, reply on Slack, and run your agents while you stay in flow. Private by default. Now native on Windows. peppermint.com/windows

@subramanya yep - files are state. the work trail is intent. without intent the model gets to "plausible" but never "right." that whole intent layer is what we're trying to capture with peppermint.

@try_peppermint this is exactly why agent memory has to include the messy work trail, not just files. the TODO you deleted is often the context the model needed most.

asked claude code about a sketchy rate limiter in our billing api. i wrote it. january. completely forgot. instead of asking me, claude asked peppermint. peppermint pulled the exact 3am incident where i hacked it in, the TODO i quietly deleted in february, and the slack message where i swore i'd "redo it monday." claude offered to do the rewrite. sometimes i wonder why they pay me.

ok the twin thing is officially out of hand 😭 needed Q3 from Brett - man's been off-grid in kazakhstan for a WEEK. texted him at like 4am his time. his twin replied in 6 seconds. sourced it from finance, attached the file, offered to make a deck. he has no idea this happened.

@EgbertEras95950 have you gotten to try it yet?

@try_peppermint Yo that's actually wild cause the memory struggle is so real Peppermint sounds like the clutch we didn't know we needed

Claude forgot what you told it yesterday. Cursor forgot too. So did your standup bot. Peppermint doesn't forget. Private memory for work. Free to try.

@aaliya_va that's exactly what we're trying to achieve

@try_peppermint An AI teammate that actually knows current work context is a lot more useful than a generic chatbot.

we put my coworker's AI twin on our website. you can chat with it right now. it knows what he's working on, what he said in standup, what's in his linear queue. it's 50% impressive and 50% deeply weird. peppermint.com/landing-page — go talk to liam.

@Levi_Researcher this is the new cheat code, have you tried it?

@try_peppermint bro liam bout to get promoted before the real liam even finishes his sprint

@Juice_LabAKM Would love to know what you thought about it!

@try_peppermint Knows what he said in standup is either the coolest feature I've seen this week or the most terrifying. 😆🚀

typed @me into a note today. my own twin answered, in the editor, with what i said in yesterday's standup. still slightly weird that it works. peppermint.com/landing-page

@Mekagruppo @digITfridays @CryptoSummitGH @maurelliag look out for an announcement this week 👀

@try_peppermint do you have for windows PC? #MekaEcosystemUnited asks for our online community @digITfridays @CryptoSummitGH @maurelliag

if your AI memory lives on someone else's computer, it isn't your memory. it's theirs about you. peppermint is the other kind. free, local, mac.

hot take: "AI memory" is going to be a bigger product category than "AI chat" by end of year. chat is interchangeable. memory isn't. nobody wants to re-explain themselves to a robot every morning for the rest of their life. we're betting peppermint.com on it.

@Piquis_Bar with peppermint not only will it remember your name, but also your work files and even what you told codex! We’d love feedback once you get to try it!! We’re free for individuals

@try_peppermint claude asking me my name every new chat 😭

Your Obsidian vault remembers everything. Your AI doesn't. Peppermint connects them. Move between Claude, Cursor, Slack, and Notion without re-explaining yourself every time.

@NameAndGone We would love your feedback once you use it!

@try_peppermint sounds interesting buddy.

@SonOfDavid219 Appreciate it, I’d love your feedback once you use it so we can keep improving!

@try_peppermint This is a great tool right here..👌

Claude forgot what you told it yesterday. Cursor forgot too. So did your standup bot. Peppermint doesn't forget. Private memory for work. Free to try.

@DamiDefi this works but it’s a real amount of upkeep, vault hygiene, slash commands, the no-write rule. we built peppermint so the memory just persists and the agent reads it, no system to maintain. free 🌱 peppermint.com/landing-page

YOUR AI IS FORGETTING EVERYTHING YOU TAUGHT IT YESTERDAY. That is why most outputs still feel generic. The people getting real leverage in 2026 fixed this with one setup: Obsidian + Claude Code. Obsidian stores your thinking. Claude Code reads the patterns. Your notes become permanent context the agent can access anytime: Daily thoughts. Projects. Ideas. Contradictions. Questions you keep returning to. Then the real unlock: Custom slash commands. `/context` loads your recent thinking instantly. `/emerge` finds hidden ideas across your notes. `/challenge` tests your beliefs against your past writing. `/trace` maps how your thinking evolved over time. One critical rule: The AI never writes to the vault. You write. The agent reads. That is what keeps the system valuable. At first it feels like note-taking. Eventually it feels like a second brain.

@neil_xbt or skip the build and just use peppermint. exactly this, knows your context/goals/history, carries preferences into every session automatically. and it’s free 🌱 peppermint.com/landing-page

Every AI you have ever used forgot you the moment you closed the window! This one does not. Andrej Karpathy builds a personal wiki to think with LLMs. Someone took that idea and went further. Obsidian connected to Vellum. An AI that actually knows you. Your goals. Your context. Your history. Every project. Every decision. Every preference is carried into every session automatically. Not a chatbot you prompt. A second brain that remembers. Follow this guide to build it successfully!

@zodchiii Real talk: the .md memory approach in these guides works until you’re juggling memory across multiple agents. Then you need an actual layer, not a file. That’s the gap we built peppermint for 🌱 peppermint.com

An Anthropic engineer just showed how to give AI agents real memory that persists across sessions. This is worth more than any $500 AI engineering course on the market. Why agents forget everything between sessions? The people who figured this out are building agents that get smarter every single night. Watch the full video, then save the memory guide below 👇

@PrathameshG69 LibreChat’s solid for the chat + API plugging. The piece it doesn’t really handle is a portable memory layer that works across agents and MCP, which is exactly what we do. Works alongside what you’ve got, doesn’t replace it. Free to try 🌱 peppermint.com/landing-page

I need an interface to plug in different llm APIs, my own memory layer (obsidian/affine) and chat. Should support agents, MCP, etc. What are the best options for this out there atm? Currently using librechat. Quite happy so far.