autobot-server

autobot-server

High-performance persistent sandbox server for the Autobot AI agent framework.

Linux only - Works with bubblewrap for 15x faster file operations.

Performance

Platform Mode Per Operation
Linux autobot (default) ~50ms
Linux autobot + autobot-server ~3ms (15x faster) ✅
macOS autobot (default) ~50ms
macOS autobot-server Not applicable (Docker overhead)

Why Linux Only?

  • Linux + bubblewrap is very fast (~5ms overhead)
    • autobot-server reduces this to ~3ms → big win
  • macOS/Windows + Docker is slower (~50ms overhead)
    • autobot-server would save ~2ms → negligible gain
    • Added complexity not worth it

Installation (Linux Only)

Linux AMD64

curl -L https://github.com/crystal-autobot/sandbox-server/releases/latest/download/autobot-server-linux-amd64 \
  -o /usr/local/bin/autobot-server
chmod +x /usr/local/bin/autobot-server

Linux ARM64

curl -L https://github.com/crystal-autobot/sandbox-server/releases/latest/download/autobot-server-linux-arm64 \
  -o /usr/local/bin/autobot-server
chmod +x /usr/local/bin/autobot-server

Usage

On Linux, autobot automatically detects and uses autobot-server if installed:

$ autobot agent  # On Linux with autobot-server installed

✓ Sandbox: bubblewrap (Linux namespaces)
→ Sandbox mode: autobot-server (persistent, ~3ms/op)

Manual usage (for testing):

autobot-server /tmp/socket.sock /path/to/workspace

Protocol

JSON over Unix socket.

Request Format

{
  "id": "req-1",
  "op": "read_file|write_file|list_dir|exec",
  "path": "/path/to/file",
  "content": "file content",
  "command": "shell command",
  "timeout": 60
}

Response Format

{
  "id": "req-1",
  "status": "ok|error",
  "data": "result data",
  "error": "error message",
  "exit_code": 0
}

Development

# Build
make build

# Build release binaries for all platforms
make release-all

# Clean
make clean

License

MIT

Repository

autobot-server

Owner
Statistic
  • 0
  • 0
  • 0
  • 0
  • 0
  • about 6 hours ago
  • February 15, 2026
License

MIT License

Links
Synced at

Sun, 15 Feb 2026 13:32:41 GMT

Languages