The useful split in this category is not features, it is where the work runs and who can read it. Everything else (how the panes are arranged, which agents are wired up) moves in a release. That does not.
| Question | Topics | Conductor | Devin Desktop | Zed | Cline | Claude Code alone |
|---|---|---|---|---|---|---|
| Licence | MIT, open source | Not stated | Not stated | Open source | Open source | Proprietary CLI |
| Platforms | macOS, Windows, Linux · PWA on phone | "Local workspaces on your Mac" | "Download for MacOS", no other build offered | macOS, Linux, Windows | VS Code extension and a CLI, not a desktop app | Any terminal |
| Price | $0 for the whole app | Free $0 · Pro $50/mo · Teams $60/mo/user | Not stated on the product page | Personal $0 forever · Pro $10/mo · Business $30/seat/mo | Free for individuals · Enterprise on request | Your Anthropic plan or API key |
| Where the work runs | Your machine | Local, or their cloud on Pro and above | "Fleets of local and cloud agents" | Your machine | Your machine | Your machine |
| Account required | None | Required for the paid plans | Sign-up required | Not for the free plan | Not for the open-source extension | Anthropic account |
| Agents it drives | Claude Code, Codex, OpenCode, Gemini CLI, any ACP agent, Anthropic and OpenAI APIs | "Bring your own subscriptions and keys" | Agent Client Protocol (ACP) | "Unlimited use with your API keys or external agents like Claude Agent, Codex CLI, and more" | Bring your own key across many providers | Itself |
Where the work runs is the only line that matters
Conductor states the trade-off more plainly than we could, so here it is in their words. Asked whether they can see your data in a cloud workspace, their pricing FAQ answers:
"Yes. Unlike with local Conductor sessions, chat messages sent in cloud workspaces are stored on Conductor's servers."
That is not a criticism. It is the honest consequence of running an agent in a sandbox you do not own, and they document it clearly. Their local mode carries the opposite promise: "Conductor never accesses your coding inputs / outputs."
The reason it is the dividing line is that it decides what you can use the thing for. Cloud workspaces keep working when you close the laptop, which local ones cannot. Local workspaces cannot leak a client's codebase to a third party, which cloud ones can be made to but have to be trusted about. Pick the constraint you actually have.
Topics has nothing to disclose here, which is a statement about its architecture and not about its virtue: there is no server of ours in the path, so there is nothing for us to store. The corollary is that if you close the laptop, your agents stop.
Choose something else if
Conductor: if you want agents to keep working after you close the lid
Their cloud workspaces run in Vercel sandboxes (8 cores, 16 GB, us-east-1, per their FAQ) and survive the app being closed. Topics has no equivalent and none is running today. Conductor is also further along on multiplayer. Note that as of 4 August 2026 their pricing page describes local workspaces as being "on your Mac".
Devin Desktop: if you want the IDE and the agents in one window
It ships a full editor around the agent surface, and it is built on the Agent Client Protocol. If your objection to a separate workspace is that you would rather not leave your editor, that objection is exactly what it is for. As of 4 August 2026 the download button offers macOS only.
Zed: if what you actually want is a fast editor
Zed is a code editor first, open source, and its free tier explicitly covers "unlimited use with your API keys or external agents like Claude Agent, Codex CLI, and more". If you spend your day writing code by hand and want an agent nearby, that is a better shape than a workspace built around watching several agents at once.
Cline: if you live in VS Code
It is an extension and a CLI rather than an application, open source, free for individuals, and its own FAQ answers "Is there vendor lock-in?" with "No!" and a list of eleven providers. Nothing to install alongside your editor because it is inside it.
Claude Code on its own: if you run one agent at a time
This is the honest one. If you start one session, watch it, and finish, a workspace is overhead. Everything Topics adds is in service of the case where there are three or four of them and you have lost track of which is waiting on you. Below that number, the terminal is fine and you should keep using it.
What Topics does not have
No hosted anything: no relay, no managed remote access, no off-machine backup. No accounts, no team roles, no authentication layer at all. If you expose it over a tunnel you have to put your own auth in front of it. Mobile is a PWA, not a native app. It is not an editor, and it is two months old.
Method
Every cell about a product other than ours is quoted or paraphrased from that vendor's own public page, fetched on 4 August 2026: conductor.build/pricing, devin.ai/desktop, zed.dev/pricing, cline.bot/pricing. Where a page does not state something, the table says "not stated" rather than filling it in.
Limits
Prices and platform support change without notice, and at least two of these products have been renamed or repositioned in the past year. This is a snapshot, not a standing fact. There are no performance numbers in the table because we have not measured anyone else's, and publishing a benchmark of a competitor we ran ourselves would deserve exactly the scepticism it would get. Feature depth is not compared at all: it moves too fast to be true for long, and it is the part you should check yourself.