Show HN: AgentDrive – persistent, versioned file storage for AI agents (tokencanopy.com)

2 points by tokencanopy ↗ HN
We’re opening the private beta for AgentDrive, a persistent file/workspace layer for AI agents.

AgentDrive gives agents:

- Versioned artifacts and folders - Hosted MCP access for coding agents - Direct upload sessions for larger files - Share links and controlled public publishing - Workspace-scoped authorization and auditability - Many supported file formats such as MD, HTML, JSON, images, videos, and even spreadsheets

The MCP surface is intentionally narrow and reviewed rather than exposing the entire storage API.

It’s a private beta, so some features are deliberately out of scope for now.

See more details: https://tokencanopy.com/products/agentdrive

Try the beta: https://app.tokencanopy.com/drive/

For MCP clients:

   claude mcp add --transport http --scope user agentdrive https://drive.mcp.tokencanopy.com/mcp
   claude mcp login agentdrive

   codex mcp add agentdrive --url https://drive.mcp.tokencanopy.com/mcp
   codex mcp login agentdrive
A public API/SDK integration experience is on the roadmap so that you can use AgentDrive in your product, but it is not available as a beta user interface today.

We’d especially appreciate feedback on:

1. Was connecting and using AgentDrive through MCP clear? 2. Which tools or storage primitives are missing? 3. What would be the first workflow you ask your agents do with AgentDrive?

Open source repo is also on the roadmap so that you can host it on your own infrastructure! Sign up at tokencanopy.com for more future updates.

Any feedback or feature requests can be sent to support@team.tokencanopy.com

I'd love to personally connect with more founders and developers who are building in the coding agent infra space: https://github.com/jiashuoz

7 comments

[ 2.2 ms ] story [ 27.3 ms ] thread
This seems especially useful when agents need to work across multiple repositories and share context or artifacts between them. Really interesting idea.
Thanks!! We also have some interesting use cases such as powering coding agents working on different hardware platforms and share files, documents, or reports in a central drive.
Also noticed you are a developer/founder from Japan! I'm visiting Tokyo this week!!!
A few questions - Are these read only? how do you relay ownership? If a rogue agent goes and deletes files for example, how would the system treats that error?

I recently open sourced Recall https://github.com/Polign/recall, which is sharing the memory and typed correctness across agents.

AgentDrive provides upload APIs so any agents can upload files to the drive.

Everything is a soft delete. Every edit and update results in a new snapshot of the file (version).

Your project looks cool! AgentDrive itself isn't a memory system - it's a cloud-based file system that can be managed by agents. In other words, you can build a memory system on top of AgentDrive.

Open source is also on the roadmap!

Thanks for the reply. would that be against some GDPR policy, not that should be a concern in the beginning, but something to think about 'wipeouts'.

I struggled with deletion and clean ups a lot and found that is the hardest part to solve of it all.

we support garbage collection jobs that periodically scan and delete objects; we will also provide an option for users to delete the files in trash, which results in permanent deletion. This is fairly common in file systems and mostly an already solved problem.