Skip to content
MCP server

fuast

Let your AI agent build, deploy and share small team apps, no account needed to start

Status and history

Answered our latest tool-list check 21 hours ago with 20 tools. It will be marked verified after 3 successful checks in 7 days.

Pricing
Free
Platforms
Web, API, CLI
Runs
Remotely
Version
2.3.0
Listing updated
3 days ago

About fuast

fuast is a free MCP server for developer tools. It runs as a hosted remote server, so there is nothing to install locally.

Apps start as two files, a typed server and a screen, on top of a runtime that already has the database, sign-in, storage and schedules. fuast type-checks the app and puts it online, and the agent tries it and reads the logs before reporting success. You then receive a link to the live app plus a claim link; opening the claim link and signing in makes the app yours to share.

One command, curl -fsSL fuast.com/install | sh, wires the MCP server and a fuast skill into the agents it finds. Any MCP client can instead use https://fuast.com/mcp, and the site has a markdown page written for agents to read.

The pricing page says it is free during alpha, with unlimited apps and members and enforced limits per app: 100,000 requests and 20,000 mutations per UTC day, 100 emails per day, and a 64 MiB SQLite database. Uploads are capped at 25 MiB a file and 1 GiB per workspace, and executables, HTML and JavaScript uploads are refused.

Install

Run in a terminal

claude mcp add --transport http fuast https://fuast.com/mcp

Add to Settings → Connectors → Add custom connector

https://fuast.com/mcp
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "fuast": {
      "url": "https://fuast.com/mcp"
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "fuast": {
      "type": "http",
      "url": "https://fuast.com/mcp"
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "fuast": {
      "serverUrl": "https://fuast.com/mcp"
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "fuast": {
      "type": "http",
      "url": "https://fuast.com/mcp"
    }
  }
}

Tools this server provides (20)

Read directly from the server.

fuast_guide
Call this first. Returns how to build, check, deploy and share a fuast app with these tools, and the rules a fuast app must follow (file layout, server and client APIs, database, auth, UI). fuast apps are small full-stack TypeScript apps called capsules: server/index.ts and client/index.tsx, no npm packages.
fuast_docs
Read one docs page as markdown (page), or search every page for words (search). With neither, lists the pages. Useful pages: capsule-api (every server and client API), database, auth, limits, storage, schedules, email, teams, examples/todo.
list_templates
Lists the working starter apps a new fuast app can begin from, with what each one demonstrates. Get a template's files with get_template.
get_template
Returns the source files of one starter template as {path: content}: a complete app that passes check_app and can be sent to deploy_app as is. Change the files to build what the person asked for.
check_app
Builds the files, type checks them and loads the server in a sandbox, without deploying anything. Returns ok with the queries, mutations, tables and tools the app declares, or ok false with diagnostics that carry file, line and message. Call it before deploy_app and fix every diagnostic.
deploy_app
Deploys the files and returns the app's URL. With a connected fuast account it creates a new private app in the person's workspace, or updates the app named in app. Without an account it deploys a temporary preview and returns preview_url, claim_url and edit_token: give the person claim_url so they can make the app theirs, and pass edit_token to later calls to keep working on it. Send every file e
get_app_source
Returns the source of the app's live deploy as {path: content}, with the deploy id to pass to update_app_files as expected_deploy_id. Large apps return the file list only: ask again with paths for the files you need. Secrets (.env files, env values) are never returned.
update_app_files
Redeploys the app with a few files changed: files maps each path to its new content, or to null to delete it. Every other file stays as it is. expected_deploy_id must be the app's live deploy (from get_app_source or the last deploy result); if someone deployed in between, nothing changes and you are told to read the source again. Builds and type checks like deploy_app.
list_apps
Lists the fuast apps the connected account can use: its own, its workspace's, and apps shared with it, with URL, sharing mode and whether the account can manage each one.
get_app
Returns one app: URL, status, sharing, handlers, recent deploys with their ids (for rollback_app), today's usage and limits, and the names (never the values) of its env vars.
get_logs
Returns recent log entries of an app, oldest first: what its handlers wrote with ctx.log, handler errors with stack traces, and platform events such as deploys. Pass the returned cursor as since to read only newer entries.
screenshot_app
Opens one page of a deployed app in a browser and returns a picture of it, with any errors the page reported while it loaded. Use it after a deploy to see what a person will see, which rows and logs cannot show: a blank page, a broken layout, text that does not fit. A preview deployed without an account is shown as a guest. An app in the account is shown signed in as the connected person. Only for
inspect_app
Reads one thing about a deployed app. manifest: handlers, schema, indexes, endpoints, read metrics. tables: each table with its fields and row count. rows: rows of one table (give table; at most 100 per call, continue with cursor). usage: today's counters and limits. schedules: cron schedules and their last runs. jobs: scheduled jobs. Only the app's owner and workspace admins can inspect.
call_app
Runs one query, mutation or action of a deployed app and returns its result and the log lines it wrote, so you can test what you deployed. It runs as the connected person (as: "me", the default), never as anyone else; as: "guest" runs it as a guest visitor and only works on public apps. On a preview deployed without an account every call runs as one guest. Mutations and actions change real data. O
share_app
Changes who can open an app. visibility: private (the owner and invited people), workspace (everyone at the owner's company), or public (anyone with the link, guests included). invite adds people by email and sends each an email; remove takes people off the list. Do this only when the person asks for it. Returns the resulting sharing.
set_env
Sets or removes environment variables of an app (read in handlers as ctx.env.NAME). They take effect at once, without a deploy. Values are write-only: no tool returns them, only the names.
rollback_app
Makes an earlier deploy of the app live again. deploy_id comes from get_app; without it, the most recent earlier deploy that is ready. The app's data stays as it is; if that deploy's schema cannot hold the current data, nothing changes.
delete_app
Deletes an app and everything in it: its data, stored files, deploys, schedules and custom domains. This cannot be undone. Only the app's owner can do it, and confirm must repeat the app's slug. Do this only when the person explicitly asks for the app to be deleted.
get_claim_link
For a preview that deploy_app deployed without an account: returns the claim link to give to the person, the preview link, and when the preview is deleted if nobody claims it. Opening the claim link and signing in makes the app theirs.
whoami
Returns the fuast account this connection acts as: email, workspace, role, and what the connection may do. If no account is connected yet, calling it starts sign-in: call it when the person has a fuast account or wants apps deployed into their own workspace instead of as a temporary preview.

What it does

Two-file apps
Each app is a typed server plus a screen, with database, sign-in, storage and schedules built in.
Type-check and deploy
fuast checks types and deploys, and every deploy is immutable and can be rolled back.
Claim link handover
You get a live link and a claim link that makes the app yours once you sign in.
Sharing modes
Apps can be private, invite-only, workspace-wide or public.
Documented limits
Per-app quotas return a 429 with a fuast_quota_exceeded error so agents can plan around them.
Self-hostable
Every limit is an environment variable for self-hosted setups.

Good for

  • Build a lunch-order app for a team
  • Put a small internal tracker online with a shareable link
  • Give a coding agent a place to deploy a quick prototype
  • Share realtime team tables across several apps

Pricing

Free

Free during alpha; paid per-workspace plans planned.

Prices change. Check fuast.com before you buy.

fuast FAQ

Do I need an account?

Not to start. Nothing needs an account until you want to keep the app, which you do by opening the claim link.

What does it cost?

Free while in alpha; paid plans are planned per workspace and alpha workspaces keep the current limits.

Which agents work with it?

Claude Code, Claude, ChatGPT, Cursor, VS Code, Codex and Gemini CLI, or any MCP client.

Is fuast free?

Yes. fuast is free to use.

How do I add fuast to Claude Code?

Run this command in your terminal: claude mcp add --transport http fuast https://fuast.com/mcp

What are the best alternatives to fuast?

The closest alternatives to fuast are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

Alternatives to fuast

  1. REA

    MCP server

    CLI and MCP server that lets coding agents reverse engineer apps and binaries

    Runs locally
    • Open source
    • 464K GitHub stars
    • Developer tools
  2. Codebase Memory

    MCP server

    Local MCP server that builds a code knowledge graph for AI coding agents

    Runs locally
    • Open source
    • 48.7K GitHub stars
    • Knowledge and memory
  3. GitHub

    MCP server

    Official MCP server connecting AI assistants to GitHub repos, issues and workflows

    Not checked yet
    • Remote
    • Open source
    • 34.5K GitHub stars
    • Developer tools
  4. MCP toolkit giving coding agents symbol-level code retrieval and editing

    Runs locally
    • Open source
    • 31.5K GitHub stars
    • Developer tools
  5. XERJ

    MCP server

    Local search engine for AI agents with cited answers over your code and docs

    Runs locally
    • Open source
    • 19.2K GitHub stars
    • Knowledge and memory
  6. Ouroboros

    MCP server

    Spec-first layer that pins requirements and verifies results of AI coding agent work

    Runs locally
    • Open source
    • 6.4K GitHub stars
    • Developer tools