Knowledge & Search
How AI searches your workspace's docs, files, and connected sources to ground its answers and the docs it writes.
Specify searches workspace documents and uploaded material as evidence for answers and writing. Building an answer from retrieved material is called RAG (retrieval-augmented generation).
Search sources
| Source | How it is searched |
|---|---|
| Workspace documents | Searches saved document content and titles. |
| Uploaded files | Searches extracted, indexed content. Availability depends on processing status. |
| Standards library | Searches published standard clauses. See Standards library. |
| External services | Connected services such as GitHub, Notion, and Slack are searched through connectors rather than copied into the workspace index. |
Semantic and text search
Search can find related meanings as well as exact terms and strings. Use document browsing for a complete listing, then open the source to read its content.
Through MCP, use knowledge_search, knowledge_grep, search_files, list_files, and read_file. External service search uses search_connectors.
Indexing and freshness
Documents and uploads become searchable when indexing completes. To find material immediately after saving or uploading, check the file listing and original content first. If it is missing from search results, check processing status and try another query.
Permissions and evidence
Search operates within the current workspace and access permissions. Items excluded from MCP exposure are also excluded from external agent access.
Open source links to check freshness and whether a citation actually supports the claim. Identify the sources needed for important decisions and read the originals to confirm the scope of your review.