Voyager
An agentic file browser to navigate your work context ✴️
490 followers
An agentic file browser to navigate your work context ✴️
490 followers
✴️ Voyager brings AI agents into your everyday file work on macOS. Find the files you need and work with AI agents using them as context, all within a familiar file browser. Meet a file browser built for finding your files and putting them to work with AI.
This is the 2nd launch from Voyager. View more
Voyager
Launched this week
Voyager brings AI agents into your everyday file work on macOS. Find the files you need and work with AI agents using them as context, all within a familiar file browser. Meet a file browser built for finding your files and putting them to work with AI.
Free
Launch Team
Render Sandboxes: a computer for every agentRun untrusted code safely. Idle sandboxes cost nothing
Promoted
Hey Product Hunt!
I'm Jongmin, co-founder of Voyager.
Voyager is an agentic file browser for knowledge workers on macOS.
As we hand more work to AI agents, it becomes more important to keep track of what each task was based on and which results we chose to use. Files and outputs pile up, but when the context stays in chat threads and our heads, we end up explaining the same background again for the next task.
We want people and AI to be able to build on the files, decisions, and results of previous work. That's why we're starting with a file browser, where the source files and outputs already are. We're building a way to work with agents there and preserve the context created along the way.
✴️ Here's what you can do in Voyager today:
File browsing: Browse and manage your files as you would in Finder.
Agent chat: Bring the agent you already use into your file browser. Voyager currently supports Claude Code, Codex, and Hermes, and we're working on support for more agents.
Tab workspace: Use Arc-like tabs to keep frequently used folders open, and organize your work and chats into tab groups.
All current Voyager features are free.
If you're looking for a better file browser, give it a try.
@jongminpark Nice work, Jongmin the point about re-explaining the same background to agents for every task is real and tying context to files is a clean answer. does Voyager keep a record of which files a task used and which result you picked?
@rodrigueshema Appreciate it! For now, there’s no system that tracks which files the agent changed. But we keep seeing people want to revert agent work when it goes wrong, so we’re researching the right way to handle it.
Picking up the second half of this. When it tries a few approaches, does it note which result I kept in the end? That's the part I'd want to check the next morning.
@0x47xavier Good question. If your agent’s memory feature is turned on, that part gets recorded on the agent’s side. Voyager doesn’t have its own memory layer yet.
Got it. So if I swap agents halfway through, the progress stays with the old one. Even a plain text export of where things stand would cover most of that for me.
@jongminpark Nice work and congrats on launch. Been using it for last few days and can't be more happy. What is on your roadmap?
@nicklaunches So glad you’re enjoying it! Beyond file browsing, we’re building a workspace for handling parallel agents better. We’re also exploring a knowledge graph for relationships between files, and cloud storage support.
i do a lot of video editing with agents, how can this help me?
@shawn_esquivel Depends on the tools you use. My take: the helpful part is organizing and keeping the work going without bouncing between your video folder and the terminal. I actually used it while making our launch video, managing all the versions we generated. Pretty useful.
can it read my notes folder on its own?
@leoheo Yep. Open a chat in the folder and send, and an agent session runs in that chat. It’s simple, same as using Claude Code. It just runs right in the folder, no switching.
@jongminpark does the chat stay in that folder for next time?
This clicks for me — the part about keeping track of what each agent task was based on, and which results you chose to use. I hand a lot of work to agents now, and the hardest part isn't getting the output, it's coming back later and re-explaining the same background because the context stayed in a chat thread. Bringing Claude Code into the file browser itself, where the source files and outputs already live, feels like the right place for that context to live. Curious how the file-as-context works in practice — does the agent see just the selected files, or the surrounding folder too?
@jcbuilds The agent session opens with the current folder as its working folder. It works with the files in that folder, same as using Claude Code. You can point it at specific files in the chat too.
claude code, codex, hermes. cursor too?
@mostafahosseini if you want Cursor, we’re ready to add it quickly.
how trustworthy it is in terms of it's not gonna delete my whole device :D
Starting with the file browser makes sense when the source files and outputs are already there. Keeping the files behind a result visible would help someone pick up the next task without reconstructing the chat.
@oleksii_sekundant
Thanks! That's exactly the problem we started from.