DoneThat vs Sidekick

The short answer

Sidekick is a free, MIT-licensed macOS app that lets you chat with a local model over your own files. It is well made, actively developed, and costs nothing.

It is also not a work tracker, and the honest version of this page says so early. Sidekick knows about the documents you point it at. It has no idea how you spent your Tuesday.

What Sidekick gets right

It is genuinely free and genuinely open. MIT licence, 3.3k GitHub stars, 591 commits and recent activity. This is a maintained project, not an abandoned experiment.

Local by default. Inference runs through llama.cpp on your own machine, so conversations stay offline. That is a stronger privacy position than any cloud assistant, ours included.

Retrieval over your own material. Files, folders and websites, with no extra software to install — which is more friction removed than it sounds.

A surprisingly full feature set for a free app: RAG, web search, function calling, deep research, memory, image generation.

An escape hatch. You can optionally connect OpenAI-compatible providers — OpenAI, Anthropic, Google AI Studio, DeepSeek, Groq — when a local model is not enough, without giving up the local-first default.

Why this is not really a comparison

Sidekick's memory is of what you gave it. Documents, folders, conversations. That is a useful thing to have, and it is entirely different from a record of your work.

DoneThat captures automatically and reconstructs the day. Nobody has to remember to file anything, and the record includes the parts you would never have thought to save — which is usually where the answer is when you cannot account for a week.

Concretely, Sidekick will not tell you that nine hours went into a recurring meeting, that your stated priority received 15% of your attention, or that you have drifted three weeks running. And it does not act: no goals, no nudges, no coaching.

Don does. It nudges you onto goals, blocks calendar time for them, flags low-value work while you can still fix it, checks your mood, reminds you of meetings, suggests breaks.

They combine well

This is the interesting part rather than a courtesy.

A local assistant is only as good as the context it can reach. Sidekick reaches the files you point it at. DoneThat exposes your reconstructed work history through an API and an MCP server — so an assistant can reason over what you actually did, not just what you filed.

If you like the local-first approach, that combination gets you a private assistant with a real record of your work behind it, rather than one that only knows your documents.

On capture: screenshots are never stored — processed in real time, immediately deleted, only the derived structured activity persists. Excluded apps are redacted on your machine before anything is sent. With your own Gemini or OpenAI-compatible model, raw activity never reaches our servers. The desktop app is open source.

Platform support

Intel Macs are out for Sidekick.

Where Sidekick is the better choice

• You want a free local assistant. Nothing to pay, nothing to sign up for.
• Nothing may leave your machine. Local inference by default.
• You want to chat with your own documents. That is precisely the job.
• You want to read the code. Fully open source under MIT.
• You have Apple Silicon and enough RAM to run a decent local model.

Where DoneThat is the better choice

• You want to know where your time went. Sidekick does not capture it.
• You use Windows or Linux, or an Intel Mac.
• You want coaching, not a chat window.
• You want goals tracked and drift noticed.
• You want an assistant that knows your work, not just your files — that is what MCP over captured history provides.
• You want peer accountability. Follow others, share summaries, Slack and Discord, accountability groups — all opt-in.

How we compared these

Everything stated about Sidekick comes from its public GitHub repository, read on 11 August 2026 and linked below. Claims about DoneThat are first-hand: we build it and charge for it.

Sidekick is free, fully open source and runs entirely locally — all advantages over a paid cloud-assisted product. We have not framed it as a rival, because it is not one: it is a local assistant, and the two are more useful together than either is as a substitute for the other.

Questions people ask about Sidekick

Is Sidekick free?

Yes — free and open source under the MIT licence, with no subscription. Local inference runs through llama.cpp so the only cost is your hardware. If you optionally connect a remote provider, you pay that provider directly.

What hardware does Sidekick need?

macOS on Apple Silicon with at least 8 GB RAM. Intel Macs are not supported, and local model quality scales with the machine you run it on.

Does Sidekick track my work?

No. It is a local AI assistant that can retrieve from files, folders and websites you point it at. It does not capture activity, reconstruct your day, or measure where your time went — that is a different category of tool.

Sidekick vs DoneThat?

They are complementary rather than competing. Sidekick gives you a private local assistant over documents you choose. DoneThat captures and reconstructs your working day and exposes that history through MCP so an assistant can reason over what you actually did, not just what you filed.