Farmscape runs complex garden and operations work that lives in a large service manual. Staff needed answers in the tools they already use, not another portal to hunt through.
The problem
Procedures for garden management, food safety, and day-to-day operations sat in a long Google Sites service manual. Employees asked the same technical and HR questions in chat, and finding the right page took time.
What we built
Endertech built Farmbot, an internal AI assistant for Farmscape employees in Google Workspace.
Knowledge base from the operating manuals. We imported the Farmscape Service Manual into a document CMS, cleaned the markdown, preserved category structure, and kept links back to the original manual paths.
Search for retrieval. Documents are indexed so the assistant can find the right sections by question, not by browsing a table of contents.
MCP server as the tool bridge. A Model Context Protocol gateway exposes search and full-document lookup. The AI agent calls those tools instead of inventing answers from general training data.
Google Chat delivery. Employees ask Farmbot in Google Chat. Dialogflow CX orchestrates the conversation, calls the MCP tools, and returns a grounded reply with citations to the source manuals.
How a question is answered
An employee asks a question in Google Chat.
The agent decides it needs company knowledge and calls the MCP search tool.
Search returns ranked manual sections, with snippets and paths to the original pages.
When needed, the agent fetches the full document for grounding.
The employee gets an answer based on Farmscape's own procedures, with a path back to the source.
What Farmscape can maintain
A Next.js CMS and Symfony API let the team manage documents, categories, and images. When manuals change, the index updates so Farmbot stays aligned with current procedures.
Outcome
Farmscape staff can ask operational questions inside Google Workspace and get answers drawn from their service manuals, with MCP-backed retrieval instead of an ungrounded chatbot.