Vector database
Coming soon. The vector database lets your agents search your data by meaning. Ask "Which customers complained about delivery?", and an agent finds the records that say it in other words, without reading the whole space.
How will it work?
You switch on vectorization for a space. AgentOS turns the text of each record into a vector, a list of numbers that captures what the text means. Records that mean similar things get vectors close to each other, so a question finds them even when no word matches.
- Text is embedded. A space's Text and Long text (Wysiwyg) fields carry the meaning. Numbers, dates and choices stay as they are, and you narrow by them with filters.
- Always up to date. A record is embedded as it is saved, so new and changed records are searchable at once.
- Same access. An agent finds only records in the modules it may read.
What will my agents do with it?
| Ask an agent | What it does |
|---|---|
| "Find the customers who complained about delivery." | Finds the records closest in meaning to the question. |
| "Which products in this space don't fit the others?" | Finds the outliers, the records furthest from a description. |
| "Sort these new products into our categories." | Finds the most similar products you already sorted and takes their category. |
| "Find prospects who look like our best customers." | Ranks records by how much they resemble the customers who bought. |
Each search can start from a filter or a segment, so it ranks only the records that already fit. Your AI tools connected over Connect AI (MCP) get the same search.
See the Roadmap for what else is coming.
Next: Import and export.
