A: Hey Never57,

That's a very insightful question about leveraging external knowledge bases when your in-platform storage might be a constraint!

Yes, it is conceptually possible to connect to a custom/external knowledge base through an MCP connector, but it depends on a few things:

1. The External Knowledge Base Must Have an API:
2. Your custom knowledge base (wherever it's hosted – another SaaS, your own database, a vector store like Pinecone, Weaviate, etc.) needs to have an API that allows for querying or searching its content.
3. An MCP Server Acting as a Bridge:
You (or a third-party, or potentially the knowledge base provider themselves in the future) would need to create an MCP server that acts as an intermediary.
This MCP server would:
- Expose MCP-compliant actions (e.g., "SearchKnowledgeBase," "GetDocumentSnippet").
- When AgenticFlow calls one of these actions (e.g., Agent says, "Use the CustomKB tool to search for 'onboarding process'"), the MCP server translates that into an API call to your actual external knowledge base.
- It then receives the results from your external KB, formats them as needed, and returns them to AgenticFlow in the MCP standard.

4. Connecting the MCP Server to AgenticFlow:
You would then add the URL of this custom MCP server into your AgenticFlow workspace (Settings > MCP Servers).
Your AgenticFlow agents can then be equipped with the "tools" (actions) exposed by this custom MCP server.

5. How it Helps with Storage Limits:
By doing this, the actual storage of the bulk knowledge resides externally.
AgenticFlow's role becomes orchestrating the query to that external knowledge via the MCP, and then using the retrieved results in its LLM prompts for RAG or other tasks.
Your AgenticFlow knowledge storage limit wouldn't be consumed by the content of that external knowledge base itself, only by any metadata or connection details for the MCP.

Current State & Effort:
- Feasibility: Technically feasible if the external KB has an API and an MCP server is built.
- Effort: Building a custom MCP server requires development effort. However, with emerging tools and platforms supporting MCP server creation (like Vercel, Cloudflare Workers, or open-source MCP libraries), this is becoming more accessible.

AgenticFlow's Role: We provide the client-side capability to connect to any compliant MCP server.

Alternative (Simpler, if direct API access is easy):
- If your external knowledge base has a simple, direct REST API for searching, you could also use AgenticFlow's API Call node within a workflow to query it directly, without needing a full MCP server in between. The workflow would fetch the data, and then an LLM node could process it.

In short: Yes, connecting to a custom/external knowledge base via a custom-built MCP connector is a valid way to work around in-platform storage limits and leverage existing data repositories. It effectively allows AgenticFlow to query external "brains." The main lift is creating that MCP server bridge if one doesn't already exist for your specific KB.

This is a great strategy for scaling knowledge access!
— Sean