AI Transparency

We use AI tools extensively in building iagent-stack. This page documents how, where, and with what guardrails so developers evaluating the SDK can understand what is and is not machine-generated.

What AI touches

Code generation scaffolding. Initial boilerplate for product client modules, test fixtures, and documentation strings may be generated with AI assistance. All generated code is read, edited, and reviewed by a human before commit.
Documentation drafting. This site, the SDK README, and llms.txt are written collaboratively with AI tooling. All copy is human-reviewed for accuracy, tone, and factual correctness before publication.
Research and analysis. Open-standards review, prior-art analysis, and competitive landscape research are accelerated with AI but cross-checked against primary sources.

What AI does not touch

Cryptographic material. Private keys, signing operations, and on-chain transactions are never generated by AI.
Product availability claims. The Shipping/Roadmap/Parked status of each product is set by a human based on actual build state, never by an AI inference.
Legal and financial disclosures. Entity identifiers, tax registrations, and formal regulatory filings are prepared by humans and reviewed where necessary by professional counsel.

Honest availability

Every product's agent-card.json declares status as live, pending, or parked. The SDK raises ProductNotAvailableError when a roadmap product is called. We do not pretend capabilities exist before they do.

Report concerns

If you believe iagent-stack or any stack product misrepresents capabilities or status, contact hello@iagentstack.com. We correct disclosed inaccuracies within 72 hours of confirmation.