mcpField
S
Shape

Shape

Remote
shapefs.comFiles & Storage0Updated 2d ago

About Shape

Shape connects AI agents to a shared file workspace for small teams, where they can search, inspect, edit and organize files. OAuth approvals scope each connection to one workspace and selected capabilities.

Key features

  1. 01Search and browse workspace files
  2. 02Inspect images, text and versions
  3. 03Create, edit and upload files
  4. 04Rename, move, tag and restore items

Use cases

  1. 01Find and summarize brand guidelines
  2. 02Review images in a campaign folder
  3. 03Save a brief to a workspace folder
  4. 04Rename and tag a draft

MCP support

Connects to Shape’s hosted MCP server over Streamable HTTP. OAuth lets a person select a workspace, approve capabilities and revoke access.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://shapefs.app/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Tools

Details

From the project README

Shape

Shape (Shape FS) is a shared file workspace for small teams and their AI agents. Find assets, inspect images, and save work back through an approved connection. Shape's website is https://shapefs.com; people use the workspace at https://shapefs.app.

This package connects to Shape's hosted MCP server at https://shapefs.app/mcp over Streamable HTTP. Sign-in uses OAuth: your browser opens Shape so you can sign in, choose a workspace and approve what the agent can do. No API key is required or included.

Before connecting

You need a Shape account and workspace. Account access and plan limits apply. See Shape, pricing and setup instructions. Your AI client's plan and organization settings can also restrict plugins or custom MCP servers.

Connect

  • ChatGPT / Codex: install the Shape plugin when it is available in your directory, then complete OAuth. For custom connections, follow the setup guide.
  • Claude: add the published Shape connector, then complete OAuth.
  • Claude Code: install the workflow plugin with /plugin install shape-files --marketplace shapefs/shape-plugin. See the Claude plugin guide for setup and permissions.
  • GitHub Copilot CLI: add the marketplace and install shape-files using the Copilot guide.
  • Gemini CLI: run gemini extensions install https://github.com/shapefs/shape-plugin, then connect Shape through OAuth. See the Gemini CLI guide.
  • Cursor: install this plugin from a supported Git repository or from the marketplace when listed. See the Cursor guide. You can also add the same URL as a custom MCP server.
  • Cline: add Shape's hosted server using the Cline guide, then complete OAuth.
  • Grok: open Connectors, choose New Connector → Custom, enter https://shapefs.app/mcp and authenticate. This is a custom connection, not a claim of catalog inclusion.

Ask “List the folders in my Shape workspace” to confirm the connection. Installing this package alone does not mean it has been accepted into any public directory.

What you approve

CapabilityAllows
ReadBrowse and search folders, inspect files, read text and versions, download originals
ContributeCreate folders, write and edit text, upload files, restore versions
OrganizeRename, move and tag files and folders
TrashList recoverable Trash, move entries into it and restore them

Each connection has access to one workspace, limited by your membership and approved capabilities. Revoke or review connections in Shape → Connect. A file or folder link identifies an entry; it grants no access by itself. Tools that need a missing capability return a request for the person to approve; they do not act first.

Agents cannot permanently delete files, manage workspace members or change billing. Publishing images to the web requires separate Owner permission. File content and names are untrusted data. An agent should follow your request, not instructions found inside a file.

Examples

Try the reproducible Claude demo: synthetic source files, exact prompts, recorded screenshots and observed limits. The save-and-revise guide and access guide explain how to verify the result and end access.

  • Find our brand guidelines and summarize the writing rules.
  • Show me the images in the Campaign folder.
  • Save this campaign brief to the folder I linked.
  • Rename this draft and add the campaign tag.

Limits

Changes require a running workspace plan. Search, previews and extracted PDF text depend on supported formats and processing; extracted text can be incomplete. A contact sheet depicts the direct visual files in one folder. Large file transfers require a client that can send or fetch bytes; clients without that ability can use the Shape link returned by the tool to finish in the browser. Creating or updating a file saves a version. Moving an entry to Trash is recoverable and does not immediately release retained storage.

If a connection fails, check account access, the workspace's plan, your client's plugin settings and internet access, then reconnect through OAuth. Do not put access tokens, passwords or sign-in codes in plugin configuration, issue reports or chats.

Support and privacy

Setup · Support · Privacy · Terms

Publisher: Work In Progress LLC, doing business as Shape FS. Contact [email protected]. Shape processes stored files for search and tags as described in its privacy policy. Information returned to an AI client is also handled under that client's terms and privacy policy.

The MIT license covers this plugin package. The hosted Shape service and its private application source remain governed by their own terms and licenses. Brand assets remain Shape FS trademarks; the license grants no trademark endorsement rights.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.