Case study
Enterprise RAG and agent knowledge platform
Designed a knowledge platform that turns internal documents, code, and operational knowledge into governed AI workflows, combining RAG, MCP tools, LangGraph orchestration, and LangSmith observability.
- RAG
- MCP
- LangGraph
- LangSmith
Context
A complex enterprise environment needed AI assistance over internal knowledge without treating a prototype chatbot as the final architecture.
Challenge
Useful information lived across internal portals, documents, and code repositories. Search alone was not enough, but a generic assistant could not be trusted without source grounding, scoped tools, evaluation, and operational visibility.
Constraints
- multiple internal knowledge sources with different structure and freshness
- enterprise authentication and access boundaries
- answers that needed traceable sources rather than plausible summaries
- agent workflows that had to be observable, testable, and recoverable
Approach
- Designed ingestion and retrieval as a governed pipeline rather than a one-off vector import.
- Separated knowledge access into focused MCP tools so agents could search, retrieve, and inspect context through explicit interfaces.
- Used LangGraph to orchestrate multi-step reasoning and tool use while keeping workflow boundaries visible.
- Added LangSmith-style tracing and evaluation practices so prompts, retrieval, and agent behaviour could be improved with evidence.
Architecture
A governed knowledge ingestion layer feeds retrieval services and MCP tools, which are orchestrated by LangGraph agents behind an API with tracing, evaluation, and operational observability through LangSmith.
In short: It only answers from your own documents, shows you where every answer came from, and asks before doing anything that matters.
- Your documentsThe manuals, contracts and files you already have.
- OrganisedEverything is split into small pieces that are easy to search.
- Right pieces foundFor each question, the most relevant passages are picked out.
- Answer writtenThe AI writes an answer using only those passages.
- Checked twiceRules check the answer; anything risky waits for a person.
- Answer with its sourceYou get the answer and a link to where it came from.
Key decisions
- RAG treated as a production knowledge system, not a chatbot feature
- MCP tools as the boundary between agents and enterprise data
- LangGraph orchestration with LangSmith observability for behaviour that can be inspected and improved
Details are anonymised. Additional context is available under NDA.
Related expertise
Curious what AI could actually do for your business?
Bring your questions, including the ones that feel too basic. In 30 minutes we go through how you work today, pick the task with the most to gain, and sketch what testing it would involve. No pitch, no obligation.
Prefer email? hello@nxtinno.com