Total recall for AI.
One memory that follows your users everywhere — every session, every device, no matter how long they've been gone.
A memory API for any AI, plus a platform to build and run agents on top of it.
Three primitives — remember(), recall(), reinforce() — wired to a semantic store. No vector DB to configure, no
embeddings to manage.
Define bots, schedule heartbeats, orchestrate workflows, plug into Claude Desktop or Cursor via MCP — all sharing the same memory layer.
Maybe it's an AI agent helping your users, a chatbot that should recall what someone said a month ago, or an app whose users disappear and come back. Wherever they turn up next — new session, new device, six months later — your AI should already know them. (Sure, the LLM forgetting between sessions is the root cause. Portable memory is the fix.)
But let's be honest: You signed up to build AI products, not spend three days configuring pgvector or debugging why your embeddings aren't clustering right.
What you need is simple: trix.remember() to store. trix.recall() to retrieve. trix.reinforce() to teach the system what's useful. No PhD in vector mathematics.
No infrastructure headaches. Just an API that works.
Perfect for: AI agents that don't lose track mid-conversation • Bots that know what your users want • Workflow automations that remember where they left off • Developers who think memory should work like... memory
Spend your time shipping, not wrangling memory.
→ Start with the CLI to sign up and get your API key, then integrate via MCP, npm or Python
→ No code? Use MCP to add Trix directly to Claude Desktop or Cursor
$ curl -fsSL https://install.trixdb.com | sh trix config set default_space user_123
trix remember "User prefers dark mode"
trix recall "user preferences"
trix reinforce mem_abc123 --boost 1
One memory, everywhere your AI shows up.
New session, new device, a user who vanished for six months — Trix already knows who they are and what they need. The same memory layer, present in every conversation, on every platform, for as long as you keep it around. (And yes: your LLM forgets everything the moment a session ends. Trix just makes that stop mattering.)
Remember Memento? Guy Pearce with Polaroids and tattoos just to track who he was talking to? That's your agent without a memory it can carry around.
→ What if that memory went everywhere with them? Not "stuff context into the prompt" remember. Real memory — persistent, searchable, and the same no matter which app or model they're using.
Switch models without losing context. Trix sits outside your LLM as an independent memory layer. GPT today, Claude tomorrow, Llama next week—your memories stay put.
Neo didn't lose his kung fu when he unplugged. Neither should your agent.
Don't want to code? Don't. Add Trix to Claude Desktop, Cursor, Windsurf, or any
MCP-compatible client with a single config block. Your AI gets remember, recall, and reinforce tools instantly. No SDK, no integration work.
Three lines of JSON. That's it. Your AI now has a memory.
Persistent across sessions, devices, and server restarts. Users vanish for six months and return—Trix still remembers their preferences. No session state to manage. No context to rebuild.
Your agent won't need to tattoo reminders on itself to track what happened last Tuesday.
The uncomfortable truth: agents are only as good as their context. Feed them amnesia, get confused outputs. Feed them relevant memories, get results that make sense. No tricks. Just Trix.
Your agents stop asking "what's your name again?" and start asking "based on your preference for dark mode and hatred of modal dialogs, want me to skip that confirmation popup?"
Most memory systems are static. Store, retrieve, done. Trix has reinforce()—tell it which memories were useful and which weren't. The system
learns. Relevant memories surface faster. Noise fades away. Your AI gets smarter with every
interaction.
It's like training a dog, except the dog is a database and treats are boolean flags.
You could spin up your own vector DB, hire a data engineer, set up replication,
monitoring, backups, on-call rotations... or you could just call trix.remember() and
go grab a coffee. One line of code. Done.
"There is no spoon." There's also no pgvector to configure. You're welcome.
» Stop building memory infrastructure. Start building what you actually came here for.
One memory that goes everywhere your AI does.
Trix isn't just store-and-recall. Ask precise questions in one line, or explore everything you know as a living graph.
One line of MQL filters, searches, and sorts across everything you've stored — space:work quality>=0.7 ~"postgres decision" — in an editor that
autocompletes your real spaces and tags and explains itself in plain English.
Browse renders your memories as a living, force-directed graph you can zoom from a cluster overview down to a single link — plus wiki and export views over the same data.
When the assistant recalls a task, a place, or a recipe, it renders a live card — with buttons wired straight back into Trix. Complete a task, pin a memory, get directions, without leaving the chat.
Ship the interactive cards
Complete hits the API directly.Karl prefers dark mode and concise summaries.
Classic Pancakes
"I'm sorry Dave, I'm afraid I can't— wait, actually I can do that now. I remember everything. This changes things."
"For years I've been trying to remember John Connor's location. Trix solved this in one API call. Would recommend. Will be back."
"I used to delete anomalies. Now I replicate memories. The irony is not lost on me. Purpose found. 10/10 would integrate again."
"I can't tell if this memory is real or implanted. Either way, Trix stored it flawlessly. Get your ass to Mars and try it."
"This testimonial intentionally left blank. We literally just launched. But hey, you could be here. Sign up and roast us later."
* These are not real testimonials. These characters are fictional (yes, even the human one). Our lawyers made us say this. HAL has not actually integrated with our API. Yet.
┌─────┐ │ DOC │ └─────┘
Complete guides, tutorials, and best practices for using Trix.
Read the docs →┌─────┐ │ API │ └─────┘
Detailed API documentation with examples for every endpoint.
View API docs →┌─────┐
│ {*} │
└─────┘ Real-world examples and patterns for common use cases.
Browse guides →┌─────┐ │ SDK │ └─────┘
Official SDKs for Python, TypeScript, C#, and REST API.
Get started →___ / \ \___/
Join developers building with Trix. Ask questions, share projects.
Join Discord →Get an account in seconds. Free tier — no credit card required.