Built-in generative tools at the point of review
Reviewers can invoke first-party generative tools from the surfaces where they already judge work—such as bound blocks, decision rails, and iterate-with-AI flows. The platform can produce candidate text, code, images, video, or document-oriented output for preview, not as a finished release.
Those candidates remain proposals. Accepting or applying a result should still flow through your review and approval patterns so experts stay accountable for what moves forward.
Custom and configurable tools fit your workflows
Beyond built-ins, teams can wire workspace tools, registered executors, or API-backed configurations so the right model or service runs for a given template, section, or block binding.
Configuration stays under your governance: which tools appear, which bindings are allowed, and which output kinds a reviewer may request—so experimentation does not sprawl into unreviewed production paths.
Chat assistants and knowledge retrieval—for conversation, not block replacement
Conversational assistants can run on a review template with LLM-backed orchestration, giving reviewers a guided way to ask questions, compare options, or sanity-check context alongside the formal review surface.
When enabled, vector-backed knowledge retrieval can enrich chat with relevant snippets from approved sources. That retrieval is scoped to the assistant conversation—not a substitute for generative replace on blocks—so chat context and structured review decisions stay clearly separated.