Persistent file storage
for AI agents.

Whimsy is a Model Context Protocol (MCP) server for file storage: it exposes your multi-cloud Whimsy library to Claude, Cursor, Codex, and any MCP-aware agent through a single endpoint, with per-agent permission scopes and a full audit log. Agents get persistent file access and a searchable knowledge index across Dropbox, OneDrive, S3, and the rest — without you handing them the keys to your whole library.

1. Frame 01

   Agentic file access, scoped

   An MCP server that exposes your Whimsy library to Claude, Cursor, Codex, and any MCP-aware agent — read, write, search files across every connected cloud, with per-tool permission scopes.

2. Frame 02

   Agents never see what they shouldn't

   OAuth-style scopes per agent, audit trail on every call, and revocable tokens. You decide which folders, providers, and operations a given agent can touch — and pull the plug instantly.

3. Frame 03

   Persistent context across sessions

   Agents lose state every conversation. Whimsy gives them durable file storage and a searchable knowledge index they can re-attach to, so a follow-up session starts where the last one ended.

The problem

Agents need file context — but not your whole disk.

Questions, answered

What is an MCP server for file storage?

An MCP (Model Context Protocol) server exposes a defined set of tools and resources — in this case, file read, write, and search — to MCP-compatible AI clients like Claude, Cursor, and Codex. Instead of pasting context into a chat, the agent calls the server directly to fetch the files it needs.

Which AI agents work with Whimsy's MCP server?

Any client that implements the Model Context Protocol — Claude Desktop, Cursor, Codex CLI, Cline, and the growing list of agent frameworks that speak MCP. You point the client at your Whimsy MCP endpoint and authorize it once.

Does the agent get access to every file I have?

Only what you scope it to. Each connected agent gets its own permission scope — specific providers, specific folders, specific operations (read, write, search) — and you can revoke its token instantly. Every call is logged.

How is this different from a local filesystem MCP server?

Local-FS MCP servers expose only the machine the agent is running on. Whimsy's MCP server spans every cloud and drive you've connected — Dropbox, OneDrive, S3, R2, NAS, your computer — behind one endpoint, so the agent can work across your real storage instead of a single laptop folder.

How does Whimsy give agents persistent context across sessions?

Whimsy keeps a durable, searchable index of your files. An agent re-attaching in a new session can search the index, re-load the documents it needs, and pick up where the last session left off — without you re-pasting context.

Where do the files actually live?

On the storage you already own — Dropbox, OneDrive, S3, your NAS, your computer. Whimsy is a control and access layer, not a file host. Files are never stored on our servers.

Is the connection between the agent and the MCP server secure?

Yes — TLS in transit, OAuth-style scoped tokens per agent, and AES-256-GCM encryption for stored credentials. Whimsy never logs file contents, and every agent call is recorded with a timestamp, scope, and operation.

Can I use this for agentic coding workflows over a Postgres or S3 source?

Yes — connect Postgres-backed storage or any S3-compatible bucket alongside your other providers. The MCP server exposes them through the same scoped, audited endpoint as everything else.