Learn
What Is a Shared Drive for AI Agents?

A shared drive for AI agents is a cloud folder that a team and its AI agents use as one shared workspace. People and agents read, search and edit the same files, so plans, notes and results are always in one current place. Unlike a drive designed only for people, it lets agents search every file with code-style tools, refuses edits based on outdated versions so agents don't overwrite each other, and gives each agent access to only the folders and actions it needs, for a limited time.
The idea comes from how agents already work. Coding agents and assistants such as Claude Code, Codex, Claude and ChatGPT keep their work in files: they write plans to Markdown, search folders for context and save their results. On one laptop that works well, because every agent shares the same disk. A shared drive for agents gives a whole team that same shared disk.
Why teams need one
When one person runs several agents on one computer, the agents coordinate through the local folder without anyone designing it. One agent writes a plan, another reads it, a third checks the output.
As soon as a second person joins, each person's agents work on their own copies of the files. The work splits into versions, and people become the connection between agents: copying the latest plan into a message, downloading a file, uploading a changed one. A shared drive for agents puts everyone's agents back on the same files. Your AI agents need a shared drive, not a group chat makes the full argument.
How it differs from a regular shared drive
Dropbox, Google Drive and similar services were designed for people. A person opens a few files, edits one at a time and saves occasionally. Their desktop apps typically sync copies of files to each computer, and their search is designed for people typing a few words. That design works well for its original users. Agents use a folder differently, in four main ways.
How many users there are. A team of five might run dozens of agents, and a larger company might run thousands. A single folder can have many agents reading and writing at the same moment.
How they search. Agents search the way programmers do: grep for an exact phrase, use a regular expression, narrow the search to one folder, then read the lines around a match. They repeat this many times in quick succession. Running those searches file by file over a network is slow, so a drive built for agents runs them on the server across every file at once.
How they write. People rarely edit the same file at the same moment, so many systems let the latest save win. Agents overlap often. A drive built for agents checks which version each edit was based on and refuses an edit made to an outdated version, so the agent re-reads the file and reapplies its change.
How they get access. People sign in with their own accounts. Agents often run in temporary sandboxes executing code that nobody has reviewed. They need access to one folder, for specific actions, for as long as the job takes, rather than a long-lived key to everything.
For a closer look at the comparison, see can AI agents use Google Drive or Dropbox?
What to look for
If you're choosing a shared drive for agents, these are the capabilities that matter most.
Ways for agents to connect. Agents work in different environments. Look for a drive that mounts as a normal folder for agents on your computer, plus an MCP server, a command-line tool and an API for agents and scripts without a filesystem. How AI agents connect to a shared drive explains when to use each.
Scoped, temporary permissions. You should be able to give an agent one folder, choose whether it can only read or also write and delete, and set the access to end on its own. Revoking one agent's access shouldn't affect any other agent. Why AI agent access should expire explains why.
Server-side search. Agents should be able to grep, use regular expressions and run full-text search across the whole drive without downloading files first.
Safe concurrent edits. Edits based on an older version should be refused, not silently applied over someone else's work, and not saved as a duplicate copy. When two AI agents edit the same file walks through a real example.
Partial reads of large files. An agent should be able to read part of a big file, such as a section of a log, a slice of a large CSV or a segment of video, without fetching the whole thing.
Protected instruction files. Files like CLAUDE.md, AGENTS.md and SKILL.md tell every agent how to behave. Agents shouldn't be able to change them, even with write access, so one compromised agent can't redirect the rest. See your agent instruction files are a security risk.
An activity log. You should be able to see what each agent read, created, changed and deleted, and when.
Safety nets. Agents make mistakes. Trash with restore and version history mean a wrong delete or a bad edit can be undone. What happens when an AI agent deletes the wrong file covers this.
Pricing that fits agents. If a drive charges per seat and counts agents as seats, costs grow with every agent you run. Pricing per person keeps that predictable.
Who it's for
People using more than one agent. If you use Claude on your phone, a coding agent on your laptop and ChatGPT in the browser, a shared drive lets all three work from the same files. See your agents, across all your devices.
Small teams and co-founders. When each person prefers a different agent, the shared drive is what keeps their work in one place. Two founders, two agents, one plan shows how this plays out.
Agencies and client work. Giving each agent access to one client's folder keeps client files apart. AI agents for agencies covers the setup.
Research and data teams. Many agents can work on the same set of interviews, papers or transcripts at once, each writing its results to the shared folder, as described in run a research project with a hundred agents.
Video and media teams. Agents can work on large footage libraries by reading only the parts they need. See AI agents for video production.
Companies running agents at scale. When many agents run in sandboxes, narrow, expiring access and protected instruction files limit the damage any one job can do.
A shared drive for agents in practice
A typical day might look like this. In the morning, a person opens NEXT.md, the handoff note an agent left the night before, and asks their agent to continue. The agent searches the project folder for earlier decisions in decisions.md, updates the plan and saves it. A teammate's agent, working in another city, picks up the plan an hour later and adds a section. Because each edit is checked against the latest version, neither agent overwrites the other.
That afternoon, a batch of agents processes interview transcripts, each with read access to the transcripts folder and write access to a results folder. Their results appear in the shared drive as soon as they're written, ready for the agent that writes the summary. When the batch finishes, their access expires on its own. How to hand off work between AI agents and share agent skills with your team cover two of the habits involved.
Fabric as a shared drive for AI agents
Fabric works as a shared drive for a team and the agents it runs.
The Fabric desktop app mounts Fabric as a drive on Mac, Windows and Linux. Files take no space on your computer and stream when they're opened, so agents on your machine work with Fabric like any other folder. Agents without a filesystem use Fabric's MCP server, which works with Claude, ChatGPT, Claude Code, Codex, Cursor and VS Code, or the CLI and API.
Fabric runs grep, regular-expression and full-text search on the server across every file. It refuses writes based on outdated versions, lets agents read part of a large file, and protects instruction files such as CLAUDE.md, AGENTS.md and SKILL.md from changes by agents. Deleted files go to a trash you can restore from, and earlier versions are kept.
Agent Vault controls what each agent can do inside Fabric: which folder, space or project it can reach, which actions it can take, and for how long, from a single task to indefinitely. Agents ask for more access when they need it, and every action is logged. Agent Passes apply the same controls to the apps connected to Fabric. Both are on the Pro and Team plans. Fabric prices per person, so agents don't need seats.
Frequently asked questions
What is a shared drive for AI agents?
It's a cloud folder that a team and its AI agents use together. People and agents read, search and edit the same files, with search, write safety and access controls designed for agents.
Can AI agents use Google Drive or Dropbox?
Agents can work with files from these services, often by downloading and uploading them. Those services were designed around people working one file at a time, so they weren't built for many agents searching and editing the same files at once.
Why do agents need server-side search?
Agents search with tools like grep and regular expressions, many times in a row. On a cloud drive, searching file by file means a network trip for each file. Running the search on the server returns results across the whole drive quickly.
What happens if two agents edit the same file?
On a drive built for agents, an edit based on an outdated version is refused. The agent re-reads the latest version and reapplies its change, so nothing is overwritten.
How do agents connect to a shared drive?
Through a mounted folder on your computer, an MCP server, a command-line tool or an API. Fabric supports all four.
Is it safe to give an AI agent access to a shared drive?
It's safer when access is scoped and temporary. Give each agent only the folders and actions it needs, set access to expire, protect instruction files, and keep an activity log. Trash and version history let you undo mistakes.
What is the difference between agent memory and a shared drive?
Agent memory stores facts and summaries an agent has learned. A shared drive holds the actual files the work is made of. Many teams use both. See agent memory vs a shared drive.
Do I need to pay for each agent?
That depends on the provider. Fabric prices per person, so agents don't need seats of their own.
Can a solo user benefit from a shared drive for agents?
Yes. If you use different agents on different devices, a shared drive lets all of them work from the same files instead of separate copies.
Related pages
You might be interested in:

How to hand off work between AI agents

Share skills with your whole team through a shared drive

How AI agents connect to a shared drive: MCP, CLI, API or a mounted drive

Let AI agents analyze large data files without downloading them

Build a CRM your agents can use (it's just a folder)

What Is a Shared Drive for AI Agents?

What Is a Virtual Data Room?

How to let clients proof and choose photos online