🚀 K-NET: An MCP-Powered Agent Network of KnowledgeBases

🌟 Inspiration

Consultancy firms spend over 70% of their time on market research and identifying client pain-points. This process is:

Manual
Fragmented
Buried in endless email threads with Insight/Knowledge Firms

We wanted to eliminate this bottleneck by designing a platform that automates discovery, validation, and synthesis of insights across distributed knowledge-bases.

Powered by GPT-OSS, MCP (Model Context Protocol), Langchain and Vector DBs (Milvus), K-NET acts as the coordination layer that seamlessly connects consultancies with intelligence from both internal and external knowledge sources.

We qualify for 3 Categories:

Best Overall
Best Local Agent: We use GPT-OSS 20B for the KnowledgeBase agent of the Insight Firms and GPT-OSS 120B for the Coordinator Agent of the Consultancy Firm. Both are Local.
*For Humanity *: Our current MVP is an Aggregator Flow for Environmental Knowledgebases using Langraph, MCP and MilvusDB in the Background.

Note (attaching Knowledge bases)

First Step

Go to Conserve.ai:- https://better-chatbot-production.up.railway.app and create an account

Second Step

Go to Knowledge Bases and add the following MCP configurations with the names

1) Renewable-knowledgebase

{ "url": "http://54.198.41.200:8000/mcp", "type": "streamable-http" }

2) Carbon-Credits-knowledgebase

{ "url": "http://35.153.181.188:8000/mcp", "type": "streamable-http" }

Third Step

1)Click on "New Chat"

2)Then click on tools,

3)You should be able to see the knowledge bases in the bottom section.

4)Click on them to make sure they're connected

5)Type the query and press Enter.

💡 What It Does

K-NET is a knowledge orchestration platform that transforms consultancy workflows by enabling:

🏢 Consultancy Value Unlock Consultancies gain:
- Accelerated research: compress weeks of market study into hours
- 360° insights: synthesis across internal + external knowledge streams
- Validated intelligence: guardrails filter noise and enforce domain alignment
- Collaborative intelligence: multiple firm-specific knowledge-bases work in sync
📡 Integration for Knowledge-Based Firms Firms can plug in their internal Insight repositories, vector databases research archives, and BI systems as first-class nodes in K-NET. This data is Served over MCP offering full modularity across the network.
🌐 Knowledge Aggregators via MCP Beyond firm-specific systems, K-NET can act as a hub of hubs — aggregating data from multiple consultancies, industry data providers, and specialized MCP-powered connectors. This leads to our current MVP where we showcase an Aggregator for Environmental Knowledgebases.

🏗️ How We Built It

K-NET is a knowledge orchestration platform that automates consultancy workflows by enabling:

🧠 Coordinator LLM A central reasoning model (120B GPT-OSS) that orchestrates market research across both local and external KnowledgeBases.
🔗 KnowledgeBase Connectivity MCP-powered connectors integrate with:
- Vector databases
- Small LLMs (20B GPT-OSS) + vector stores
- RDBMS
- Object-based storage systems
🛡️ Guardrail LLMs Embedded safety layers that refine and validate knowledge before passing it to the Coordinator.
⚡ Automated Research & Synthesis What once took hours of email back-and-forth now executes in near real-time with richer, validated insights.

⚔️ Challenges We Faced

Retrieving, processing and Inserting Environmental Data.
Setting up the langchain <-> MCP Server and Milvus <-> MCP Server FLow.
Building stable MCP connections across heterogeneous data sources.
Ensuring guardrail models filtered without cutting valuable insights.

🏆 Accomplishments

Built a scalable+pluggable multi-agent architecture optimized for consultancies.
Cut manual research processes significantly.
Created an MCP-powered interoperable framework that works with both modern LLM/vector stores and traditional databases.

🔮 What’s Next for K-NET

🌐 Expanded MCP Ecosystem Plug-and-play connectors for Salesforce, Notion, BI tools, and more. Make the Knowledge Aggregator Agent Flow on Langgraph more robust.
🛡️ Adaptive Guardrails Domain-specific small LLMs (finance, healthcare, policy): Supervise Fine-tuning.
🔒 Secure Enterprise Deployment On-premise & hybrid-cloud options ensuring data confidentiality.

✨ With K-NET, consultancies unlock a future where AI orchestrates knowledge across systems, freeing experts to focus on strategies and social impact.