Blog
Your agents, across all your devices

Claude on your phone, a coding agent on your laptop and ChatGPT in the browser can all work from the same files once those files live in Fabric. Here's how each one connects and what that changes day to day.
If you use more than one AI tool, the simplest way to make them work together is to give them the same files. Connect Claude, ChatGPT and your coding agent to Fabric, and each of them can search, read and write the same notes and documents from whichever device you happen to be on. The phone, the laptop and the browser tab stop being three separate workspaces.
If you use several AI tools, the pattern may be familiar. A plan drafted with Claude on the phone stays in that conversation. The coding agent on the laptop knows the repository and nothing else. ChatGPT in a browser tab has its own memory of you, which neither of the others can see.
Each tool is good at its job, and each starts every session knowing only what you paste into it.
A day with three agents
Here's a plausible version of how this looks for one person, a freelance developer we'll call Sam. It's an illustration, not a customer story.
On the train in the morning, Sam opens Claude on their phone and asks it to look at clients/harbor/NEXT.md in Fabric and draft a reply to the client's latest question. Claude reads the note Sam's coding agent left the night before, sees that the export bug is fixed but untested on Windows, and writes a reply that says so. Sam tweaks two lines and sends it. Then they record a quick voice note in the Fabric app about an idea for the client's onboarding flow. Fabric transcribes it (transcription is on paid plans), and Sam moves the note into the same client folder.
At the desk, Sam starts Claude Code in the client's project. It reads NEXT.md, finds the voice note transcript in the folder, and asks whether the onboarding idea should go into this week's work or the backlog. Sam says backlog. The agent adds it to backlog.md, fixes the Windows issue, and rewrites NEXT.md before Sam stops for lunch.
In the evening, Sam opens ChatGPT in a browser to think through a proposal for a different client. They ask it to search Fabric for the rates they quoted on similar projects this year. ChatGPT finds three earlier proposals, reads the pricing sections, and helps write the new one, which Sam saves back to that client's folder.
Three agents, three devices, and no file was emailed, downloaded or pasted between them. Each agent saw the work the others had done because the work was in one place.
How each device connects
Fabric runs on the web, as a desktop app for Mac, Windows and Linux, and as iOS and Android apps, so the files themselves are available everywhere. The agents connect in one of two ways.
On your phone. Claude connects to Fabric through Fabric's MCP server, and Fabric is listed in Claude's official connector directory. Anthropic's help center says connectors set up on Claude's web or desktop app become available in the Claude apps for iOS and Android the next time you sign in, so you can connect Fabric once at your desk and use it on the phone afterwards. Anthropic describes installing connectors from mobile as being in beta, which is another reason to set it up on a computer first. The Fabric mobile app is the other half: it's where you record voice notes, save links and files through the share sheet, and back up photos and screenshots, all of which your agents can then find.
On your laptop. Coding agents have two routes. The Fabric desktop app mounts Fabric as a drive, with files that take no disk space and stream when opened, so any agent working on your computer can read and write it like a local folder. Or you add the MCP server directly: in Claude Code that's claude mcp add fabric --transport http https://mcp.api.fabric.so/mcp, and in Codex it's codex mcp add fabric --url https://mcp.api.fabric.so/mcp. Cursor and VS Code with Copilot take the same server URL in their MCP settings. Scripts can use the CLI.
In the browser. ChatGPT adds Fabric as a connector. Open Settings, go to Connectors, choose Add Connector, enter the server URL https://mcp.api.fabric.so/mcp, name it and save. In a new conversation, click More, choose Add Connectors, select Fabric and sign in. Sign-in is OAuth for every tool, so there are no API keys to copy between devices.
How AI agents connect to a shared drive compares the mounted drive, MCP, the CLI and the API in more detail.
What travels between them
The useful distinction when you set this up is between files and chat history.
Each AI tool keeps its own conversations and, in many cases, its own memory of you. Those stay inside that tool. What all three agents share is what's in Fabric: Markdown notes, documents, PDFs, spreadsheets, transcripts, saved articles. If a decision or a half-finished thought matters beyond one conversation, it belongs in a file.
Two small files do most of the work. A NEXT.md note at the end of each session says what's done and what's next, so whichever agent you open later starts from the same point. A decisions.md log records what you chose and why, a line or two at a time. Neither needs a template. How to hand off work between AI agents shows the pattern, and how to stop re-explaining context to ChatGPT covers the same problem from the chat side.
Fabric also has memories, which agents connected through MCP can create and read. They suit short facts about you and your preferences, while the files hold the work itself. Agent memory vs a shared drive explains how the two fit together, and the memory engine behind Fabric describes M-1.
When two of your agents touch the same file
Working alone doesn't mean only one agent is running. You might leave a coding agent tidying a project on the laptop while you ask Claude on your phone to add a line to the same notes file.
Fabric checks each edit against the version it was based on. If the coding agent tries to save a change to a file that Claude has updated since the agent last read it, Fabric refuses that save. The agent re-reads the file and applies its change to the current version. Both edits survive, and you don't end up with a stray copy of the file to sort out later. When two AI agents edit the same file has a step-by-step example.
The same drive handles the other things a busy agent can get wrong. Search runs on Fabric's side, including grep, regular expressions and full-text search, so an agent on a phone connection doesn't have to download your folders to look through them. A large file can be read in parts. Deleted files go to a trash you can restore from, and earlier versions of files are kept.
Keeping each agent in its lane
Connecting several agents to your files is a reason to think about what each one should reach. Agent Vault, on the Pro and Team plans, treats each external agent as a member of your workspace with its own access. You might give the coding agent read and write access to the project folders, let ChatGPT read your proposals folder and nothing else, and keep the folder with tax documents out of reach of all of them. Access can last for one task, a day, a week or indefinitely, and every read and edit is logged, so you can see which agent did what.
Instruction files such as CLAUDE.md, AGENTS.md and SKILL.md are protected: agents can't change them, even with write access. If you keep a CLAUDE.md in a project folder describing how you like your code reviewed, it stays as you wrote it.
Who this suits
This setup earns its keep for anyone who already moves between devices and tools: freelancers and consultants switching between clients, developers who plan on the phone and build on the laptop, students and researchers who capture on the go and write at a desk. If your files are already scattered, your files are on five devices covers that side of it.
Fabric prices per person, so connecting more agents doesn't add to the bill. See pricing for individuals.
The same idea scales to a team, where it matters even more. Two founders, two agents, one plan shows two people doing what Sam does alone, and your AI agents need a shared drive, not a group chat makes the broader case. If the term is new, start with what is a shared drive for AI agents?
Frequently asked questions
Can I use Claude on my phone with my files?
Yes. Fabric is in Claude's connector directory. Connect it on Claude's web or desktop app, and Anthropic says connectors carry over to the Claude iOS and Android apps the next time you sign in.
How do I use the same files in ChatGPT, Claude and Claude Code?
Connect each tool to Fabric's MCP server at https://mcp.api.fabric.so/mcp. Each one then searches and reads the same Fabric files, and can create notes, folders, tasks and tags there.
Do ChatGPT and Claude share memory?
Not directly. Each keeps its own conversations and memory. If both are connected to Fabric, they can read the same files and the same Fabric memories, which is the practical way to share context between them.
Can a coding agent on my laptop use Fabric like a normal folder?
Yes. The Fabric desktop app mounts Fabric as a drive on Mac, Windows and Linux. Files take no disk space and stream when opened, and agents read and write them like any local folder.
What happens if two of my agents edit the same file?
Fabric refuses an edit that was based on an outdated version of the file. The agent re-reads the latest version and reapplies its change, so neither edit is lost.
Can I stop an AI agent seeing certain folders?
Yes. With Agent Vault on the Pro and Team plans, you choose which folders each agent can reach, which actions it can take and how long its access lasts.
Do I pay extra for each AI tool I connect?
No. Fabric prices per person, so connected agents don't need seats of their own.
Does Fabric work on Android and Windows?
Yes. Fabric has iOS and Android apps, a desktop app for Mac, Windows and Linux, a web app, and browser extensions for Chrome, Safari and Firefox.
Related pages
Other blog posts:

Agent memory vs a shared drive

Can AI agents use Google Drive or Dropbox?

What happens when an AI agent deletes the wrong file

Why AI agents give up searching your files

Why AI agent access should expire

Your agent instruction files are a security risk

When two AI agents edit the same file

Ask your files, don't build a dashboard