The vocabulary of verifiable answers.
A short, plain-English vocabulary for grounding AI agents with VerveContext — from what provenance means to how a scoped key works. New to the idea? Start with our full explainer, What is grounding?
- Grounding
- MCP
- Citations
Answers with a source
Grounding
Anchoring a model's answer to verifiable external data fetched at answer time, so a claim traces to a real source rather than to training memory.
Source
A live, callable origin of a fact — a market feed, a weather reading, a domain record. VerveContext exposes 300+ sources as tools your agent can check.
Provenance
The origin of a fact and enough detail to check it. Provenance is what makes an answer verifiable instead of merely plausible.
Citation
The visible reference a grounded answer carries, pointing to the source that produced each claim — a receipt a person or system can follow.
Why grounding matters
Hallucination
A confident answer a model invents when it lacks the fact. Grounding is the most direct defense: hand the model the real value so it needn't guess.
RAG
Retrieval-Augmented Generation — retrieving passages from documents you ingested and embedded. Great for your own knowledge; grounding covers live external facts RAG can't hold.
Real-time
Fetched at the moment of the answer, not remembered from training. 'Right now' means right now — prices, weather, and news as they currently are.
Verifiable
A claim whose truth can be independently checked because it arrives with its source. The line between an agent that sounds right and one that can prove it.
MCP, tools & control
MCP
The Model Context Protocol — an open standard for exposing tools and data to models. VerveContext grounds agents by serving live sources as MCP tools.
Tool
How a source appears to a model over MCP: a callable capability with a model-readable name and schema the model can invoke on its own.
Scoped key
A credential that grants an agent access to only the sources you allow — so a workflow can reach exactly what it should, and nothing more.
Audit log
A record of what each agent pulled to ground an answer. It's how a team sees, after the fact, exactly which sources shaped a response.