Sandbox
@beam-cloud/airstore

Virtual filesystem and MCP tools for AI agents

Airstore turns integrations into folders and files that an agent can read with normal filesystem tools. It also provides MCP tools as executables, so the agent can search data and act on connected services from the mounted workspace.

188 stars11 forksGoUpdated 4mo ago
Airstore
Eli Mernit - Beam373 views • 7 months ago
Who it's for

Builders who want their agent to work with connected data through files and command-line tools.

What it delivers

You can give your agent scoped access to live data without re-explaining context or wiring each app by hand.

What it does

Source views

Create folders of files from natural language queries or structured filters, such as invoices, open PRs, or high-priority issues.

Integrations

Connect services like GitHub, Gmail, Google Drive, Linear, Notion, Slack, and web sources.

MCP tools as executables

Expose tools such as GitHub, Wikipedia, and filesystem actions as command-line binaries under the mounted tree.

Team workspaces

Share integrations and source views across a team.

Local mode

Run the stack on your own infrastructure with your own MCP servers and config file.

How to get it

  1. 1Run
    curl -fsSL https://get.airstore.ai | sh
  2. 2Run
    airstore login
  3. 3Run
    airstore mount ~/airstore
  4. 4Your source views are now available as local directories
    ls ~/airstore/gmail/invoices/
    # stripe-invoice-jan-28.eml
    # aws-invoice-jan-25.eml
    # digitalocean-jan-22.eml
  5. 5Run
    cat ~/airstore/gmail/invoices/stripe-invoice-jan-28.eml
    
    From: billing@stripe.com
    Subject: Your invoice from Stripe
    Date: Jan 28, 2025
    
    Amount: $249.00
    Status: Paid
  6. 6Run
    cd ~/airstore
    claude

README

Airstore

The filesystem for AI agents

License: AGPL v3 GitHub stars

Website · Docs · Demo Video · Discord


What is Airstore?

Airstore adds any source of data into a virtual filesystem. Connect Gmail, GitHub, Linear — then describe what you need in plain English. Results appear as files that Claude Code can read.

⭐ If this is useful, give us a star

Why files?

Claude Code already knows how to read files, search directories, and work with file contents. By turning your integrations into files, Claude can use the same tools it uses for code, for your data.

Source views also let you scope exactly what Claude can access. Instead of granting access to your entire inbox, create a view with just "invoices from last week" — or use structured filters for precise control.

Features

  • Source views - Natural language or structured queries that materialize as folders of files
  • Integrations - Connect GitHub, Gmail, Google Drive, Linear, Notion, Slack, Web
  • Tools - MCP servers exposed as executable binaries
  • Team workspaces - Share integrations and source views across your team
  • Local mode - Run entirely on your own infrastructure

Quick Start

1. Sign up

Go to app.airstore.ai and create an account.

2. Connect an Integration

In the dashboard, go to Settings → Integrations and connect a service (GitHub, Gmail, Drive, and more.).

3. Create a source view

Click New Folder and describe what you want in natural language, or use structured filters:

  • "Open PRs in acme/api that need review"
  • "Invoices I received in email last week"
  • "High priority issues in the current sprint"

4. Install the CLI and Login

curl -fsSL https://get.airstore.ai | sh
airstore login

5. Mount the filesystem

airstore mount ~/airstore 

Your source views are now available as local directories:

ls ~/airstore/gmail/invoices/
# stripe-invoice-jan-28.eml
# aws-invoice-jan-25.eml
# digitalocean-jan-22.eml
cat ~/airstore/gmail/invoices/stripe-invoice-jan-28.eml

From: billing@stripe.com
Subject: Your invoice from Stripe
Date: Jan 28, 2025

Amount: $249.00
Status: Paid

6. Use with Claude Code

cd ~/airstore
claude

Ask Claude to work with your data:

  • "Summarize these invoices and tell me the total"
  • "Which invoices are unpaid?"
  • "Extract vendor names and amounts into a CSV"

How it works

~/airstore/
├── linear/
│   └── design-issues/     # Source view
├── github/
│   └── open-prs/          # Source view
├── gmail/
│   └── invoices/          # Source view
└── tools/
    ├── github             # MCP tool executable
    ├── linear             # MCP tool executable
    └── gmail              # MCP tool executable

Source views contain data from your integrations, materialized as files. They sync automatically in the background, or manually on demand.

Tools are MCP servers exposed as executables that let you take actions:

~/airstore/tools/github create-issue --repo=acme/api --title="Bug fix needed"

Tools output JSON and can be piped together with standard Unix tools:

~/airstore/tools/wikipedia search "albert" | grep -i 'einstein'

Installation

The install script handles everything:

curl -fsSL https://get.airstore.ai | sh

This automatically:

  • Downloads the correct binary for your platform (macOS/Linux, Intel/ARM)
  • Installs FUSE dependencies
  • Configures permissions

Verify the installation:

airstore --version

System requirements

PlatformStatus
macOS (Apple Silicon)Supported
macOS (Intel)Supported
Linux (x86_64)Supported
Linux (ARM64)Supported
WindowsComing soon

Local Mode (Self-Hosted)

For development or self-hosted deployments, you can run Airstore entirely on your own infrastructure with your own MCP servers.

Prerequisites

FUSE (required):

# macOS
brew install fuse-t

# Ubuntu/Debian
sudo apt install fuse3

Node.js (for npx-based MCP servers):

# macOS
brew install node

# Ubuntu/Debian
sudo apt install nodejs npm

Configuration

Create a config.local.yaml file:

mode: local

gateway:
  grpc:
    port: 1993
  http:
    host: 127.0.0.1
    port: 1994

tools:
  mcp:
    filesystem:
      command: npx
      args: ["-y", "@modelcontextprotocol/server-filesystem", "/tmp", "/Users"]

    memory:
      command: npx
      args: ["-y", "@modelcontextprotocol/server-memory"]

    wikipedia:
      command: npx
      args: ["-y", "@modelcontextprotocol/server-wikipedia"]

    github:
      command: npx
      args: ["-y", "@modelcontextprotocol/server-github"]
      env:
        GITHUB_TOKEN: "${GITHUB_TOKEN}"

Mount locally

airstore mount ~/airstore --config config.local.yaml

Your tools are now available:

ls ~/airstore/tools/
# filesystem  memory  wikipedia  github

# Search Wikipedia
~/airstore/tools/wikipedia search "artificial intelligence"

# List files
~/airstore/tools/filesystem list_directory /tmp

# Get help for any tool
~/airstore/tools/filesystem --help

Building from source

git clone https://github.com/beam-cloud/airstore
cd airstore
make build
./bin/cli mount ~/airstore --config config.local.yaml

Documentation

Full documentation at docs.airstore.ai:

Contributing

PRs welcome! See CONTRIBUTING.md for guidelines.

Join our Discord to discuss ideas and get help.

License

AGPL 3.0 License - see LICENSE for details.

Files in the repo

Repository payload25 top-level entries
  • .github
  • bin
  • cmd
  • deploy
  • docker
  • docs
  • e2e
  • googleapis
  • hack
  • internal
  • manifests
  • pkg
  • proto
  • scripts
  • sdk
  • .gitignore
  • .go-version
  • .stignore
  • AGENTS.md
  • config.local.yaml
  • go.mod
  • go.sum
  • LICENSE
  • Makefile
  • README.md

Discussion (0)

Ask about usage, or say what you built with it

Sign in to join the discussion.

No comments yet. Be the first to say what this is good for.

More connectors

Real-time global intelligence dashboard. AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface

86k

High-performance code intelligence MCP server. Indexes codebases into a persistent knowledge graph — average repo in milliseconds. 158 languages, sub-ms queries, 99% fewer tokens. Single static binary, zero dependencies.

43k

Universal provider proxy for OpenAI Codex & Claude Code — use any LLM (Claude, Gemini, Grok, DeepSeek, Ollama…) with Codex CLI, App, SDK, and Claude Code

14k
okf-memory/
okf-agent-memory

Git-native persistent memory for AI coding agents. Implements Google OKF v0.2 with sub-300µs in-memory BM25 search, embedded MCP server, and progressive disclosure. Slashes token bloat by 80% with zero external databases or dependencies. Built in pure Go.

547
tirth8205/
code-review-graph

Local-first code intelligence graph for MCP and CLI. Builds a persistent map of your codebase so AI coding tools read only what matters, with benchmarked context reductions on reviews and large-repo workflows.

31k
2akouwu/
reverify

Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is the proving ground. MCP server + CLI.

1.1k