Droid Engine

An open-source, lightweight CLI AI assistant. Droid is built for speed and simplicity — minimal dependencies, fast startup, and no heavy infrastructure required. Perfect for quick tasks and lean development environments.

Last updated: 2025-01-15

Overview

Droid is an open-source CLI AI assistant designed with one principle: do more with less. Where other engines require Node.js runtimes, persistent servers, or complex authentication flows, Droid is a single binary with a straightforward config file. It starts in milliseconds and gets out of your way.

QuantumReef integrates Droid via its structured JSON output flag, piping responses directly into the session UI. Because Droid is stateless and lightweight, it is especially well-suited for CI/CD pipelines, remote SSH sessions, and resource-constrained environments like Raspberry Pi or cloud micro-VMs.

Best for quick, focused tasks

Droid excels at single-purpose tasks: generate a function, explain a snippet, draft a commit message. For complex multi-step agentic workflows, consider Goose or Claude Code instead.

Installation

Install Droid

bash

# Install via Homebrew (macOS / Linux)
brew install droid-ai/tap/droid

# Or download the binary directly
curl -fsSL https://get.droid.ai | sh

# Or build from source
git clone https://github.com/droid-ai/droid
cd droid && cargo build --release
sudo mv target/release/droid /usr/local/bin/

# Verify
droid --version

Initial configuration

bash

# Run the setup wizard
droid init

# Or manually create the config file
mkdir -p ~/.config/droid
cat > ~/.config/droid/config.toml << 'EOF'
[provider]
name = "anthropic"
api_key = "sk-ant-..."
model = "claude-haiku-3-5"

[behaviour]
max_tokens = 4096
stream = true
EOF

Enable in QuantumReef

Go to Settings → Engines → Droid and toggle on. QuantumReef locates the droid binary from your PATH. Droid's lightweight footprint means engine switching to Droid is near-instant.

Configuration

config.toml

~/.config/droid/config.toml

[provider]
name = "anthropic"          # anthropic | openai | ollama | groq
api_key = "sk-ant-..."
model = "claude-haiku-3-5"  # lightweight and fast
base_url = ""               # override for custom/self-hosted endpoints

[behaviour]
max_tokens = 4096
stream = true               # stream tokens as they arrive
system_prompt = ""          # global system prompt prepended to all requests
temperature = 0.3           # lower = more deterministic output

[tools]
allow_shell = false         # enable only if you need code execution
allow_write = false         # enable only if you need file writes
shell_timeout = 30          # seconds before shell commands time out

QuantumReef Engine Settings

Setting	Default	Description
Provider	anthropic	LLM provider: anthropic, openai, ollama, groq
Model	claude-haiku-3-5	Model — lighter models recommended for speed
Max Tokens	4096	Maximum tokens in the response
Stream	true	Stream tokens into QuantumReef as they arrive
Allow Shell	false	Enable shell execution tool (disabled by default)
Allow Write	false	Enable file write tool (disabled by default)

Use Cases

Droid works best on focused, single-purpose tasks
Use Case	Example Prompt
Commit message generation	"Write a conventional commit message for: added OAuth login with GitHub"
Code snippet explanation	"Explain this TypeScript type: Record<string, () => Promise<void>>"
Quick function drafting	"Write a debounce function in TypeScript with a configurable delay"
Shell command lookup	"What's the rsync command to sync a local folder to a remote server?"
Error diagnosis	"What does this error mean: ECONNREFUSED 127.0.0.1:5432"
Regex generation	"Write a regex to match ISO 8601 date strings"

Ollama for fully local operation

Set provider = "ollama" and point base_url at your Ollama instance for completely offline, zero-cost AI assistance. Droid's minimal resource footprint pairs perfectly with local model inference.

CI/CD Integration

Droid's stateless binary model makes it ideal for use in automated pipelines. Use it to generate PR descriptions, summarise diffs, or check code for obvious issues as part of a GitHub Actions or GitLab CI workflow.

.github/workflows/ai-review.yml

name: AI PR Summary
on: [pull_request]

jobs:
  summarise:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - name: Install Droid
        run: curl -fsSL https://get.droid.ai | sh

      - name: Generate PR summary
        env:
          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
        run: |
          git diff origin/main...HEAD |           droid "Summarise this diff as a concise PR description"           > pr-summary.txt

      - name: Post summary as PR comment
        uses: actions/github-script@v7
        with:
          script: |
            const fs = require('fs');
            const body = fs.readFileSync('pr-summary.txt', 'utf8');
            github.rest.issues.createComment({
              issue_number: context.issue.number,
              owner: context.repo.owner,
              repo: context.repo.repo,
              body
            });

When to Choose Droid

Use Case	Why Droid Excels
Quick one-off queries	No server startup, no auth flow — just run and get an answer
CI/CD pipelines	Single binary, scriptable, no Node.js or Python runtime needed
SSH / remote environments	Tiny binary works over slow connections and minimal VMs
Ollama / local models	Pairs perfectly with self-hosted inference for zero-cost usage
Resource-constrained hosts	Minimal RAM/CPU footprint compared to Node-based engines