ClawBox

Open-source cloud file system for AI agents
Semantic search • Folders • File sharing • Self-hostable

Live Demo • Self-Host Guide

---

What is ClawBox?

ClawBox is an open-source file storage platform built for AI agents. Upload files, search by meaning, organize into folders, and share with anyone — all via API or CLI.

For agents: Store files, search by meaning, organize into folders — all via API.

For humans: Upload files, get a share link, search across documents. Like a smarter Dropbox with an API-first design.

Key Features

Feature	Description
Semantic Search	Search files by meaning, not keywords. Powered by Gemini embeddings.
Multimodal	Index text, PDF, Word, Excel, PowerPoint, CSV, images, audio, video.
Virtual Folders	Organize files with paths like /docs/reports/q1.pdf.
File Sharing	Generate share links — anyone with the link can download.
Google Login	Sign in with Google for 10 GB storage (1 GB free without login).
Self-Hostable	Docker Compose and cloud-friendly deployment options for local or hosted setups.

---

Quick Start

Option 1: Docker

git clone https://github.com/Alfra-AI/clawbox.git
cd clawbox
cp .env.example .env       # Edit to add your Google API key for search
docker compose up -d
docker compose exec app alembic upgrade head

ClawBox is then available at http://localhost:8000.

Option 2: Use the hosted version

No setup needed — just visit clawbox.ink.

Option 3: CLI

pip install .
clawbox init                    # Get a token
clawbox upload report.pdf       # Upload
clawbox search "quarterly revenue"  # Semantic search

If you want to connect ClawBox to a coding or task agent, see ClawBoxSkill/SKILL.md.

---

API

All endpoints (except public ones) require Authorization: Bearer <token>.

Core

Endpoint	Method	Auth	Description
POST /get_token	POST	No	Get a free token (1 GB storage)
POST /files/upload	POST	Yes	Upload a file (with optional path for folders)
GET /files	GET	Yes	List files (filter by folder, recursive)
GET /files/{id}	GET	Yes	Download a file
PATCH /files/{id}	PATCH	Yes	Move/rename a file
DELETE /files/{id}	DELETE	Yes	Delete a file
POST /search	POST	Yes	Semantic search across files
POST /files/embed	POST	Yes	Generate/retry embeddings

Sharing

Endpoint	Method	Auth	Description
POST /files/{id}/share	POST	Yes	Create a share link
GET /s/{code}	GET	No	Download via share link

Auth

Endpoint	Method	Description
GET /auth/google	GET	Start Google OAuth login
GET /auth/me	GET	Get current user info

---

Supported File Formats

Format	Search Support
Text, Markdown, JSON, XML, CSV	Full text extraction
PDF	pdfplumber
Word (.docx)	python-docx
Excel (.xlsx)	openpyxl
PowerPoint (.pptx)	python-pptx
Images (PNG, JPEG, GIF, WebP)	Gemini multimodal + captioning
Audio, Video	Gemini multimodal embedding

---

Self-Hosting

Single Server

docker compose up -d
docker compose exec app alembic upgrade head

With MinIO (S3-Compatible Storage)

docker compose -f docker-compose.cluster.yml up -d
docker compose -f docker-compose.cluster.yml exec app alembic upgrade head

Any Cloud

ClawBox works with any S3-compatible storage. Set S3_ENDPOINT_URL:

Provider	S3_ENDPOINT_URL
AWS S3	(leave empty)
MinIO	http://minio:9000
GCS	https://storage.googleapis.com
DigitalOcean Spaces	https://{region}.digitaloceanspaces.com
Cloudflare R2	https://{account_id}.r2.cloudflarestorage.com

See Self-Hosting Guide for the full local, cluster, and cloud setup flow.

---

Configuration

Variable	Default	Description
DATABASE_URL	postgresql://...	PostgreSQL connection
STORAGE_BACKEND	local	local or s3
S3_ENDPOINT_URL	(empty)	S3-compatible endpoint
GOOGLE_API_KEY	(empty)	Gemini API key (enables search)
GOOGLE_CLIENT_ID	(empty)	Google OAuth (enables login)
SESSION_SECRET_KEY	change-me	Session signing (change in prod)
APP_URL	http://localhost:8000	Public URL for share links

See .env.example for all options.

---

Architecture

┌─────────────────────────┐
│     Web UI / CLI        │
└────────┬────────────────┘
         │ HTTP API
┌────────┴────────────────┐
│     FastAPI Server      │
│  (stateless, scalable)  │
└────────┬────────────────┘
         │
    ┌────┴────┐
    │         │
┌───┴───┐ ┌──┴──────────┐
│ PostgreSQL │ │ Object Storage │
│ (pgvector) │ │ (Local/S3/MinIO)│
└────────┘ └─────────────┘

• PostgreSQL + pgvector — metadata, users, embeddings, search
• Object storage — file content (local filesystem, S3, MinIO, GCS, etc.)
• Gemini — embeddings + multimodal indexing (optional)

---

Project Structure

src/
├── main.py           # FastAPI app, routing, middleware
├── config.py         # Settings from environment
├── models.py         # SQLAlchemy models
├── auth.py           # Bearer token authentication
├── storage.py        # Storage backend (local/S3)
├── embeddings.py     # Gemini embeddings + text extraction
├── database.py       # Database connection
├── cli.py            # CLI tool (clawbox command)
├── routes/
│   ├── tokens.py     # Token management + settings
│   ├── files.py      # File CRUD + sharing + folders
│   ├── search.py     # Semantic search
│   └── oauth.py      # Google OAuth
└── static/
    └── index.html    # Web UI

---

Contributing

Contributions are welcome! Please open an issue or submit a PR.

---

License

MIT