Run Droid locally on Mac.

Step 1 · Launch via Ollama

One command to start.

$ ollama launch droid

Step 2 · Pre-flight memory check

Will your Mac fit it?

Droid pairs best with Qwen 2.5 Coder 32B at Q4_K_M (~19.0 GB). Add ~4–8 GB on top for the agent's working set, plus headroom for the rest of your dev stack. Recommended: 48 GB Mac (minimum 32 GB).

$ devpulse ai --before-load 19456 --auto-clean
before: Won't fit — 6.4 GB short
  - unloaded idle ollama model: qwen2.5:7b (4.2 GB)
  - killed 4 zombie procs (612 MB reclaimed)
after:  Fits comfortably — 3.1 GB headroom

# safe to launch
$ ollama launch droid

Why this matters: Fully agentic + large model = real OOM risk. Pre-flight every launch.

Step 3 · Babysit long sessions

Don't lose progress to OOM mid-task.

Agent runs that span hours hit memory pressure as context grows. devpulse babysit watches and auto-cleans without crashing the session.

# in another tmux pane
$ devpulse babysit --target-free-mb 8192 --json > droid.log &

# emits NDJSON tick / cleanup events you can checkpoint on:
{"event":"tick","tickNum":47,"availableForAIMB":7200,"pressure":"free<8192MB",...}
{"event":"cleanup","reasons":"free<8192MB","reclaimedMB":5400,...}

Tips for Droid

What we've learned running it.

• Sub-agents fan out — the orchestrator + multiple workers can hold several model contexts warm. Pre-flight assuming 1.3-1.5x your model's footprint
• BYOK to Ollama/vLLM means the model runs on your hardware — `devpulse ai --before-load <MB> --auto-clean` is the right gate before kicking off a Droid task
• If you're flipping between Droid in your terminal and Droid in VS Code, the Electron-backed IDE adds 1-2 GB. Quit one or the other if RAM is tight
• Qwen 2.5 Coder 32B is the strongest local pairing for Droid's code workloads

Recommended models

Best local pairings for Droid.

Related agents