Agentic Work · Click

Every pixel's right-click is a door to an agent.

It doesn't just draft text and leave the rest to you — it finishes the job. One shortcut captures the text, file, or screen in front of you, and a full agent runtime takes it from there.

Status
Prototype
Last verified
2026-07-18
Access
TODO_CONTENT_CONFIRMATION

The problem it solves

Explaining an error message, tweaking a SQL query, or converting a file usually means switching windows, describing your request from scratch, and hand-carrying the result back. Those in-between steps cost time and force you to re-explain context again and again.

Click removes the steps in between: it carries whatever you're on straight to an agent, the work finishes there, and the result lands right back where it belongs.

How it works

  1. Select what you're on — text, a file, or a region of the screen.

  2. Press the shortcut and type what you want.

  3. Click hands the object and its access path to the agent, untouched.

  4. The result lands in the panel — one key pastes it into place.

Verified capabilities

A launcher that does the work.

It actually runs the SQL you selected, actually converts the file, actually pastes the result right where your cursor is — there's no hand-carrying step left in between.

The agent decides what it is.

Whatever you select — text, a file, a region of the screen — travels to the agent as-is, source included; you never pre-label it as SQL, a log, or an email, and if something is beyond it, it says so plainly.

Never explain the context again.

No switching windows, no describing your request from scratch — the agent already sees what you're looking at, whether that's an email reply, an error, or data inside a PDF.

Full capability, all on its own.

A computer and a Claude Code subscription are all it takes — no server, no account, no extra infrastructure; UAIS ecosystem connections are optional add-ons, and the flow is identical with them off.

Architecture and integration

Click is a desktop client with a Tauri (Rust) core, a React interface, and local SQLite storage; it runs as an overlay window triggered by a global hotkey or right-click.

It's currently built for macOS, with Windows and mobile support on the roadmap. UAIS ecosystem connections (the bus, shared services) are an optional add-on layer — with them off, it still runs at full capacity with just a computer and a Claude Code subscription.

Trust and control model

Click captures whatever you're on in its raw form and leaves the decision of what it is to the agent, rather than a local classifier guessing upfront; if a needed capability isn't installed, the flow tells you plainly instead of silently doing the wrong thing.

Core functionality runs with just a computer and a Claude Code subscription; ecosystem connections are an optional layer, never a requirement for the base flow to work.

Known limits / areas for improvement

macOS only for now

Windows and mobile support don't exist yet; they're on the roadmap without a committed date.

Access model still settling

The product is at prototype stage; its access model and general availability haven't been confirmed by the product owner yet.

Plugin coverage is limited

The capability needed for some object types may not be installed yet; when that happens the flow tells you and stops, rather than silently continuing.

Evidence

Click is currently in development/prototype stage; there is no published independent test report, benchmark, or customer case study yet. The most accurate source for a current demo and scope is a direct technical call.

Let's walk through the flow that starts on your desktop.

Request a technical call