1. What iAgent is
iAgent is a local desktop app focused on workflows in its current public scope. It is not a browser SaaS, and it is not designed around uploading all data to a hosted workspace.
Users describe work in plain language, and the app turns that request into a reviewable workflow. Existing nodes are used first; when something needs extension, the app can prepare a reviewable rule or script.
2. How data is handled
Source data such as tables, documents, and reports stays on the local machine by default. File parsing, amount handling, format conversion, and storage writes should run locally where possible.
When a user uses Web Chat, managed model credits, AI-assisted generation, semantic understanding, document Q&A, rule generation, or script drafts, related prompts, document excerpts placed into model context, table content, tool call parameters, and model responses may be sent through the cloud model path.
iAgent may use Cloudflare AI Gateway or similar cloud logging systems to store AI interaction logs for debugging, quality analysis, security review, cost control, billing, and service operations. Users should not submit data they are not allowed to process, secrets, payment information, identity numbers, or other content they do not want to enter the cloud model path.
3. Workflows and human review
iAgent workflows are not meant to let a model directly modify core records. Output from models, rules, and scripts goes through review so users can inspect the explanation, inputs, outputs, and risk notes.
High-risk actions such as external-system writes, regulated filing, system export, and file overwrite require human confirmation. Model output must not bypass review and directly write core record tables or overwrite user files.
When the app generates a script node on the spot, that script output also goes only to the review queue. Users should understand what the script is supposed to do before saving or reusing it.