Write, make, collaborate, and publish – with a personal AI that knows your projects, files, and ideas. Now available from your terminal too. A personal memory layer for you, and your agents.
We're back! And this time we built something specifically for the developers and tinkerers in the community.
Today we're launching Fabric CLI 🖥️
Your entire Fabric library, right from your terminal. Search it, save to it, talk to it. Never tab out again.
✦ What is it? A free command-line tool that connects to your Fabric workspace. One line to install, no dependencies. You get: • fabric save "revisit the auth retry logic" • fabric search "mixture of experts diagram" • fabric ask "summarize everything tagged with project-atlas"
AI search that finds what you meant in about 200ms. An AI agent that can summarize, tag, move files, create notes. All without leaving the shell.
✦ Why we built this
We kept hearing from developers: "I love Fabric but I live in my terminal." Fair enough. But there's a bigger reason too. Coding agents like Claude Code, Codex, and Cursor are incredible, but they forget everything between sessions. They have no persistent memory.
With the CLI, any agent that can run a shell command can pull context from your Fabric library before starting a task and save what it learned when it's done. Your agents get smarter over time instead of starting from scratch.
✦ Quick recap of Fabric for the uninitiated: 🧠 An AI workspace that organizes itself around how you actually think 🤖 Frontier memory engine (more on this soon) 🔍 Semantic search across everything you've ever saved 📱 Apps on web, mobile, desktop, and now the terminal
Free to use. One line install. Give it a spin and let us know what you think!
We read every piece of feedback and it genuinely shapes what we build next. Thank you for being part of this with us 🙏
I basically live in the terminal and kept alt-tabbing to the app to save things or look stuff up. now I don't. one line install, no deps, and it just works
Report
This is basically how I use Fabric now. I just pipe things into fabric save without really thinking about it. The search is surprisingly fast!
Personally, very happy that we made this – super handy to save a quick thought without leaving the terminal!
Report
How large is the memory? We’ve started using Claude Code for AI-native development - one developer can work with 10–20 Figma pages over several days to build the frontend. Ideally, it should remember the entire project and all dependencies) How does this work on your side?
Persistent memory for coding agents via a shared CLI-accessible workspace is the right move — agents forgetting context between sessions is the biggest compound-interest loss in agentic coding. `fabric save/search/ask` from any shell is the low-friction primitive agents need. Curious how Fabric handles conflict when two agents write overlapping context to the same library.
Broadly, these scenarios are handled the same way a person does naturally – so for example, updating priors if there's credible new information, lowering confidence if the information conflicts, or allowing co-existence and expanded retrieval if the context is non-conflicting and additive.
Replies
Fabric
Hey Product Hunt! Johnny here from Fabric 👋
We're back! And this time we built something specifically for the developers and tinkerers in the community.
Today we're launching Fabric CLI 🖥️
Your entire Fabric library, right from your terminal. Search it, save to it, talk to it.
Never tab out again.
✦ What is it?
A free command-line tool that connects to your Fabric workspace. One line to install, no dependencies. You get:
• fabric save "revisit the auth retry logic"
• fabric search "mixture of experts diagram"
• fabric ask "summarize everything tagged with project-atlas"
AI search that finds what you meant in about 200ms. An AI agent that can summarize, tag, move files, create notes. All without leaving the shell.
✦ Why we built this
We kept hearing from developers: "I love Fabric but I live in my terminal." Fair enough.
But there's a bigger reason too. Coding agents like Claude Code, Codex, and Cursor are incredible, but they forget everything between sessions. They have no persistent memory.
With the CLI, any agent that can run a shell command can pull context from your Fabric library before starting a task and save what it learned when it's done. Your agents get smarter over time instead of starting from scratch.
✦ Quick recap of Fabric for the uninitiated:
🧠 An AI workspace that organizes itself around how you actually think
🤖 Frontier memory engine (more on this soon)
🔍 Semantic search across everything you've ever saved
📱 Apps on web, mobile, desktop, and now the terminal
Free to use. One line install. Give it a spin and let us know what you think!
We read every piece of feedback and it genuinely shapes what we build next.
Thank you for being part of this with us 🙏
Dynbox.app
I basically live in the terminal and kept alt-tabbing to the app to save things or look stuff up. now I don't. one line install, no deps, and it just works
This is basically how I use Fabric now. I just pipe things into fabric save without really thinking about it. The search is surprisingly fast!
Fabric
Personally, very happy that we made this – super handy to save a quick thought without leaving the terminal!
How large is the memory? We’ve started using Claude Code for AI-native development - one developer can work with 10–20 Figma pages over several days to build the frontend. Ideally, it should remember the entire project and all dependencies) How does this work on your side?
Fabric
@natalia_iankovych The total memory space can be huge, as large as your Fabric storage! (Current maximum 2TB).
As for how the memory works, we've been doing novel R&D on own memory engine – which we'll share more about soon.
jared.so
Persistent memory for coding agents via a shared CLI-accessible workspace is the right move — agents forgetting context between sessions is the biggest compound-interest loss in agentic coding. `fabric save/search/ask` from any shell is the low-friction primitive agents need. Curious how Fabric handles conflict when two agents write overlapping context to the same library.
Fabric
@mcarmonas A good question!
Broadly, these scenarios are handled the same way a person does naturally – so for example, updating priors if there's credible new information, lowering confidence if the information conflicts, or allowing co-existence and expanded retrieval if the context is non-conflicting and additive.