Kit – Unified AI Workspace
I've been frustrated by the same thing for years: code in one window, docs in another, terminal in a third, and an AI assistant in a fourth tab that knows nothing about any of the others. Kit is my attempt to fix that.
It's a desktop app (Electron, macOS/Linux) where a code editor, Chromium browser, terminal, git panel, autonomous agent, visual pipeline builder, whiteboard, email client and calendar all share context. The AI in the terminal can read the page you have open in the browser. The agent uses your actual project files. The editor has one-click summarise, error check and test generation on the current file.
The agent piece is what I'm most interested in getting feedback on. You give it a task in plain English like "scaffold a REST API with auth and write tests" — and it plans and executes in a loop using read_file, write_file, list_dir, run_command and search_project. Every file write or shell command asks permission before it runs. Drop a .kitrules or AGENT.md in any project and the agent reads your conventions at the start of every task.
There's also Stairs, a visual pipeline builder for chaining AI prompts, shell commands, HTTP requests and file operations useful for things like "fetch API data → summarise with Claude → write to a daily report."
Works with OpenAI and Anthropic keys. MIT licensed.
GitHub: https://github.com/raiyanyahya/kit
Site: https://thekit.dev
You’re solving the right problem.
Most “AI dev tools” still feel like bolt-ons.
Editor plugin.
Chat tab.
Terminal wrapper.
Different surfaces pretending to be one workflow.
The real shift is shared context.
Once the browser, terminal, repo, and agent actually see the same state, the product stops being “AI for coding” and starts becoming the operating layer for technical work.
That’s the more important positioning move here.
“Unified AI workspace” is directionally right, but still undersells what’s actually happening.
This is closer to:
the environment where execution context stays intact
That matters more than convenience.
Because the pain is not tab switching.
It’s context loss between actions.
That’s the real wedge.
Also worth tightening the frame around trust:
the agent is useful because it can act
the product is usable because it asks permission
That distinction is doing more work than the feature list.
Strong build.
The positioning just needs to lean less “AI workspace” and more “execution environment with shared context.”
Also — Kit is clean, but broad enough that you may eventually want a tighter .com if this becomes the category layer.
Xevoa.com would fit that direction well.