Ground its answers

Choose what happens beyond the corpus

Trade coverage for grounding explicitly instead of leaving fallback behavior ambiguous.

Captured from the product Demo workspace

Choose strict abstention or allow a general-knowledge fallback when the corpus does not cover the answer.

What this changes for your team.

Strict mode asks the agent to abstain when attached knowledge does not support an answer. Fallback mode allows the model to continue from general knowledge but requires that source class to be disclosed, making the policy visible to authors and readers.

How it works in practice.

01
Choose strict abstention for agents that must stay within approved material.
02
Choose disclosed fallback where a useful answer beats stopping, and every one of them arrives labelled as such.
03
Review grounding metadata to see whether knowledge, tools, fallback, or no recorded source supported the reply.

What you can plan around.

The behaviour you can design against, stated concretely.

Fallback policy is an agent setting included in the published setup.

Grounding summaries record general-knowledge fallback separately from knowledge evidence.

Strict abstention is enforced around the model's output rather than requested inside the prompt, and each turn records which source class it landed on - so the policy is auditable after the fact, not only asserted before it.

Complete the operating path

View all capabilities

Ground its answersLink answers back to their sources

Let readers move from an agent's response back to the organization material that informed it.

Ground its answersKeep answers grounded as knowledge grows

Let Yekar.AI adapt retrieval automatically as the selected knowledge grows.

Build the agentBuild agents with AI

Turn a working brief into a reviewable agent draft without starting from a blank setup.

Bring one real process