New — Skrptiq drives your real apps

AI That Works the Way You Would.

Build workflows where each step is a different specialist — one drafts, another reviews, another matches your voice. Now they act in your real, logged-in apps too, and only ever do what you approve.

A run's specialist steps listed with their models, and the step-lineage graph connecting input to result in the gutter

The flagship — Connector

It runs your real apps — and only when you say so.

Skrptiq drives your real, logged-in apps — starting with your browser and Gmail. It shows you every action before it runs and does exactly what you approve. Nothing happens behind your back.

01Works in your real accountsYour live Gmail, your live tabs. No API keys to wire, no copies of your data handed over.
02Every action is shown and gatedA run lays out its steps up front and pauses at an approval break before it acts.
03It sends exactly what you approvedNo drift. If the page or task isn't what you approved, it refuses rather than guessing.

The approval card: review the proposal, then approve as-is or type an edit before the run continues

How it works — one run

One blog post. A full quality process.

A real pipeline from the hub — not a mockup. Specialist steps take turns, pause at an approval break, and score the work before it ships.

 Start→ Structure→ Content ◗ Approval break →Checks→ Output 

↓ Walk the run — 5 steps

Step 1 — Start

Set up the run.

Fill in a few typed inputs — topic, audience, voice, word count — and they map straight to the variables each step uses. Optional fields default sensibly; required ones gate the run until they're set.

Run inputs screen with typed fields for topic, audience, voice, and word count mapped to workflow variables

Step 2 — Structure

See the whole pipeline.

Every specialist step is laid out as a flow — ideation, briefing, drafting, headlines, polish, SEO, compliance. See the model each step uses, the data it passes on, and where an approval break pauses for you.

Workflow flow view showing connected specialist steps, model badges, and an approval break

Step 3 — Content

Write, polish, offer options.

Specialists take turns — one drafts the post, another matches your voice and grammar strictness, and another generates headline options to choose from: direct, question, how-to, contrarian, each with a rationale.

A specialist step producing five headline options — direct, question, how-to, list, and contrarian — each with a character count

Step 4 — Checks

Score the work.

Dedicated review steps grade the output against the plan — an SEO pass rates on-page quality, a brief-compliance check confirms it matches the spec. Open any step to see exactly what it scored and why.

The SEO review step scoring the draft — an overall SEO score of 74 out of 100 with specific fixes

Step 5 — Output

Publish-ready, fully traced.

Every step is logged — which ran, which model, what it produced. The deliverable sits alongside the inputs that made it and the full chain of specialist steps, so you can trace any line back to its source.

↑ Back up · → next section

A completed run with all specialist steps green, shown alongside the inputs and the finished deliverable

The operating layer

A personal operating layer, not a chat window.

The run is the surface. Around it: a library you own, a catalog you can pull from, and connections into the tools you already use.

Workspace object map visualising connected skills, prompts, and workflows

Hub catalog of importable, signed community skrpts organized by category

MCP server settings exposing your skrpts to clients like Claude Code and Cursor over localhost

Built-in assistant answering a question about which skrpts to use and how they fit together

Get access

Get early access.

Skrptiq is a local-first desktop app, in beta and auto-updating. Register below and we'll send you a download link.

Platform status

macOSFull

Browser connector + Gmail, the run engine, builtins & object library. Signed and notarized.

LinuxCore + connector

Browser connector, the run engine, builtins & object library.

WindowsCore

The run engine, builtins & object library.