The coding agent harness that drives its own terminal panes
Zerg AI's own coding harness, in evaluation beta. Ask for Vim or htop and the agent opens them in panes beside it. Panes can hold more agents, each on its own model. Available models depend on your sign-in route and provider access. Mac with Apple chips.
Zerg Client (the ztc command) is the coding harness we built from scratch, and we use it every day. It has two halves: a TypeScript agent that plans, edits and tests your code, and a native Rust layer that draws the screen. They talk over one protocol.
Because we own the screen, Zerg Client comes with its own multiplexer, zmux, built for the agent to drive. Ask for your file in Vim, or htop while it works, and the panes open beside it. The agent reads them and types into them, and you can watch every one.
Panes can hold agents too, each on its own model. In the video, DeepSeek V4.1 Flash writes tests in one pane while Kimi K3 writes the README in another. Sign in once with Zerg Router or bring your own OpenAI or DeepSeek key. Available models depend on your sign-in route and provider access.
You can also switch models in the same session. In one of our recorded sessions, DeepSeek V4.1 Flash built a feature and its tests passed, 13 of 13. We switched to Kimi K3 for a review. It found two things it would block a PR on, fixed both, and the tests still passed.
The agent doesn't need a screen, so the same Zerg Client runs in your terminal, in our desktop app, inside your own app, or headless on a server. We run pools of them in the cloud to build agentic products for our customers.
It can also look at what it builds, take the mouse and keyboard when you allow it, and run simulations where every agent plays a role.
The desktop beta is for Mac with Apple chips only. The interactive terminal client also needs a Mac with Apple chips; headless use runs on macOS, Linux and Windows. Download and setup: https://zergai.com/ztc
Zerg Client is licensed under the Zerg Source Available License 1.0. It is not open source. Evaluation is free, including at work; ongoing commercial use needs a separate signed license: licensing@zergai.com.
What would you have it build first?
About Zerg Client on Product Hunt
“The coding agent harness that drives its own terminal panes”
Zerg Client launched on Product Hunt on November 5th, 2026 and earned 79 upvotes and 4 comments, placing #17 on the daily leaderboard. Zerg AI's own coding harness, in evaluation beta. Ask for Vim or htop and the agent opens them in panes beside it. Panes can hold more agents, each on its own model. Available models depend on your sign-in route and provider access. Mac with Apple chips.
On the analytics side, Zerg Client competes within Software Engineering, Developer Tools and Artificial Intelligence — topics that collectively have 1M followers on Product Hunt. The dashboard above tracks how Zerg Client performed against the three products that launched closest to it on the same day.
Who hunted Zerg Client?
Zerg Client was hunted by Idan Beck. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
For a complete overview of Zerg Client including community comment highlights and product details, visit the product overview.
Idan Beck
Hi Product Hunt, I'm Idan, founder of Zerg AI.
Zerg Client (the ztc command) is the coding harness we built from scratch, and we use it every day. It has two halves: a TypeScript agent that plans, edits and tests your code, and a native Rust layer that draws the screen. They talk over one protocol.
Because we own the screen, Zerg Client comes with its own multiplexer, zmux, built for the agent to drive. Ask for your file in Vim, or htop while it works, and the panes open beside it. The agent reads them and types into them, and you can watch every one.
Panes can hold agents too, each on its own model. In the video, DeepSeek V4.1 Flash writes tests in one pane while Kimi K3 writes the README in another. Sign in once with Zerg Router or bring your own OpenAI or DeepSeek key. Available models depend on your sign-in route and provider access.
You can also switch models in the same session. In one of our recorded sessions, DeepSeek V4.1 Flash built a feature and its tests passed, 13 of 13. We switched to Kimi K3 for a review. It found two things it would block a PR on, fixed both, and the tests still passed.
The agent doesn't need a screen, so the same Zerg Client runs in your terminal, in our desktop app, inside your own app, or headless on a server. We run pools of them in the cloud to build agentic products for our customers.
It can also look at what it builds, take the mouse and keyboard when you allow it, and run simulations where every agent plays a role.
The desktop beta is for Mac with Apple chips only. The interactive terminal client also needs a Mac with Apple chips; headless use runs on macOS, Linux and Windows. Download and setup: https://zergai.com/ztc
Zerg Client is licensed under the Zerg Source Available License 1.0. It is not open source. Evaluation is free, including at work; ongoing commercial use needs a separate signed license: licensing@zergai.com.
What would you have it build first?