AI in Fenceline
Fenceline uses language models and embeddings in focused workflows. Generated output should be reviewed before it becomes customer-facing or drives a business decision.
Drafting and extraction
Language-model integrations support tasks such as drafting descriptions, structuring document content, and extracting information in application workflows. Users remain responsible for reviewing the result.
Semantic search
Fenceline stores embeddings for supported entities and uses vector search to retrieve related contractor data. Results depend on what has been indexed and on the caller's authorized tenant context.
MCP development
The repository includes an MCP server with RAG, materials, supplier, BOM, WooCommerce, and Playwright-backed browser building blocks. Hosted transports and tool dispatch are not yet end-to-end production ready.
View current MCP status →Use AI output as assistance
Verify generated measurements, pricing, permit information, material availability, and customer details against authoritative sources.
Keep human approval around updates, orders, messages, and other side effects. A tool definition or model response is not confirmation that an action completed.