Your AI-Assisted Automation app is a local-first workspace for building reliable automations from PDFs, code, and interfaces. Bring your own Claude or OpenAI provider, inspect every step, and keep every applet under your control.
YAAA ships five modular editor plugins — PDF, DXF/CAD, Code, GUI, and Spreadsheet — combined with a manifest-first applet runtime and a bring-your-own AI companion. Each surface is inspectable, capability-gated, and local-first.
Mark up, annotate, redact, and work with PDF forms and measurements. Export and sign-related workflows stay inside the editor without sending files to an external service.
Open, inspect, annotate, and export technical drawings in DXF format. Built on the shipping cad-editor plugin — the same local-first, capability-gated model as every other editor.
Author manifest-first applets in Lua, Python, or C# with a full editor for source work, diagnostics, and execution. Build, test, and version automation logic you can read.
Design applet interfaces visually with a model-tree editor, live preview, layout commands, bindings, component tokens, and diagnostics — all inside the workspace.
Open, edit, recalculate, and export XLSX and CSV files. Built on the yaaa_spreadsheet engine with full formula recalculation — running entirely on your machine.
Connect Claude or OpenAI through your own account and provider key. The AI proposes steps, explains logic, and helps refine automations — you review and approve before anything runs.
Automations execute on your local machine by default. Cloud connections, AI provider calls, license checks, and update services are explicit, documented boundaries — not hidden dependencies.
Applets run inside a capability-gated sandbox. If a workflow requests access it has not been granted, the runtime denies that action before it reaches host resources.
YAAA does not bundle or resell AI models. Connect Claude or OpenAI through your own account and credentials. The AI companion inspects applet state and editor context through app-owned controls.
Ask it to draft an applet from a plain-language description, explain a workflow step, suggest a fix for applet code, or propose a GUI layout — then review and approve before anything runs.
Live AI-driven operation of every editor surface is in active development and framed around explicit human approval — the AI co-designs; you decide what runs.
Each applet declares its identity, language, entrypoint, and requested capabilities in applet.json. Source lives in main.lua, main.py, or main.cs — readable, versioned, and inspectable at every stage.
Most automation tools run your work on someone else's servers. YAAA inverts that: the work runs on your machine, and every cloud, AI, or network connection is an explicit, inspectable choice.
| Typical cloud automation | YAAA local-first |
|---|---|
| Runs on the vendor’s servers | Runs on your machine by default |
| Execution hidden behind a dashboard | Inspect every manifest, step, and log |
| Your files uploaded to the provider | Files stay local unless you grant access |
| Model and platform lock-in | Bring your own AI provider; connections stay optional |
| Opaque, all-or-nothing permissions | Capability-gated sandbox — denied by default |
Describe the outcome you want, refine it in the editors, approve which capabilities the applet may use, then run it locally — with logs and explicit approvals for sensitive actions.
Tell the AI companion what you want to build. It proposes an applet with a manifest, declared capabilities, and initial logic you can read before anything is saved or run.
Refine the automation in the Code Editor, adjust document workflows in the PDF Editor, or prototype the interface in the GUI Designer — each editor is a separate, authorized plugin.
Review the capabilities your applet requests: file access, network access, system commands, and sensitive operations. Grant only what the workflow needs; the sandbox denies the rest.
Execute the automation on your machine. Inspect logs, review outputs, and re-run with confidence. Cloud and AI provider calls are optional and explicit — not required to run.
Ship more with fewer resources. Automate the repetitive glue work between tools, documents, and scripts — without giving up visibility into what runs.
Standardize recurring tasks without learning a full development stack. Describe the process; review the proposed steps; run it with confidence.
Capture a process once and reuse it across people and tools. Applets stay readable, versioned, and shareable — no tribal knowledge required.
A controllable AI layer on top of Lua, Python, or C# scripts and APIs. Inspect every manifest, gate every capability, and version everything with Git.
The platform is built around a secure local workspace: OAuth-backed AI provider connections, manifest-first applets, modular editor plugins, and explicit gates for sensitive actions.
Connect Claude or OpenAI and use AI as a collaborator that proposes, explains, and helps refine automations — without hiding the workflow from you.
The PDF, DXF/CAD, Code, GUI, and Spreadsheet editors are modular plugins; their tiles appear only when the matching package is installed and authorized.
Create applets from applet.json plus a Lua, Python, or C# entrypoint. Keep requested capabilities minimal until a workflow genuinely needs more access.
Packages are encrypted with AES-256-GCM, signed with Ed25519, and denied when the user, entitlement, or package integrity check fails.
Automations run on your machine by default. Cloud and provider connections are optional, explicit boundaries — not hidden requirements.
High-impact actions are framed around previews, confirmations, and audit evidence. The AI suggests; you approve what runs.
Describe the outcome, review the proposed applet, adjust the steps, and run it locally with explicit approvals for sensitive actions.
