Overview of Petavue MCP

1. About Model Context Protocol (MCP)


The Model Context Protocol (MCP) is an open standard that defines how AI models connect to external tools, data sources, and systems in a structured way.


Instead of trying to push all relevant information into a model’s short-term memory (its context window), MCP acts like a universal adapter between the model and the external world.


With MCP:

  • Models can discover which tools or APIs are available on a given server.


  • They can call those tools with structured inputs.


  • They receive structured outputs that they can reason with reliably.This avoids brittle prompt hacks or unscalable “memory stuffing,” and gives models clean, governed access to enterprise systems.

2. The 3 Core Entities of MCP


MCP is built on three key components that work together to deliver answers and insights:


  1. MCP Server


    • Hosts the actual tools and connectors — for example, Petavue’s Planner, Core Analysis, or Summary.


    • Executes the heavy lifting, such as querying data, running transformations, or applying business rules.


    • Think of it as the tool workshop + factory floor where the work gets done.


  2. MCP Client


    • The messenger or courier.


    • Connects to the MCP Server, discovers which tools are available, sends requests, and collects the results.


    • Usually an LLM (Claude, GPT, Llama, etc.). In Petavue, we’ve built our own MCP Client to manage tool execution and keep orchestration efficient.


    • It ensures communication is standardized so different models can work with the same tools.


  3. MCP Host / Orchestrator


    • The project manager and strategist.


    • Decides which tool(s) to use, in what order, and how to combine their outputs to answer a question.


    • Provides reasoning on top of the structured results returned by the Server.


    • In Petavue, we’ve built our own MCP Orchestrator as well to manage tool execution and keep orchestration efficient.

3. How MCP Works


Each time you ask Petavue a question, MCP coordinates the workflow behind the scenes.


  1. You ask a question

    • Example: “Show me the win-rate by region for Q2.”


  2. Tool discovery

    • The MCP Client connects to the MCP Server and checks which tools are available (Planner, Core Analysis, Summary, etc.).


  3. Decision-making
    • The Orchestrator (LLM) reviews your query and decides:
      • Which tools to call
      • In what sequence should they be executed

      • How to handle intermediate results


  4. Request execution

    • The MCP Client sends structured requests (e.g., JSON payloads) to the MCP Server.


  5. Tool processing
    • The MCP Server executes the requested tools — running data queries, calculations, or validations.

    • It then returns structured outputs.


  6. Assembly & reasoning

    • The Client and Orchestrator work together to interpret results, add reasoning, and assemble the final narrative.


  7. Final delivery
    • Petavue presents the insights in chat — often with charts, summaries, or action-ready recommendations.

4. Tools Exposed by Petavue MCP Server


Petavue makes a set of tools available through MCP. These tools work together to handle analysis, data exploration, visualization, and reporting without hallucinations or guesswork. You don’t need to call these tools manually; Petavue orchestrates them for you.

Tool

Purpose

Example Use
About-Petavue Provides system-level info about Petavue’s capabilities and connected data. Invoked automatically at the start of a question to check the system context.
Understand-Data Retrieves schema details: sources, tables, columns, definitions, key definitions (KDs), metrics. “What tables are available for pipeline analysis?”
Planner Generates a step-by-step analysis plan (mandatory first step for any analysis). Creating a new analysis, modifying an existing plan, or handling follow-up queries.
Execute-Plan-Steps Runs each step of a plan created by the Planner, one step at a time. Executes Step 1: filter deals, Step 2: calculate win rate.
Tool-Output Fetches outputs of completed steps or reports. Displays results back in Petavue’s UI.
Clarification-Answer Resolves ambiguities or conflicting instructions in a user query or plan. “Did you mean revenue by region or by territory?”
Summary Generates a plain-language summary of a completed analysis, with insights and recommendations. Good for executive-style reporting.
Find-Out-How Explains the methodology and logic behind a report or analysis. “How was the win-rate calculated in this report?”

Note: You don’t have to invoke these tools yourself. The Petavue MCP Client handles orchestration automatically; your role is simply to ask questions, review the generated plans, and approve execution.


4. Significance of Petavue’s MCP Client


Petavue has introduced its own MCP Client, which means it can now connect directly to the MCP server and orchestrate tools without relying on an external client like Claude or GPT.


For you as a user, this brings three key benefits:


  • Consistency → The same analysis flows whether the orchestrator is Claude, GPT, or another model.


  • Control → Petavue manages orchestration, giving more stability and fewer unexpected errors.


  • Flexibility → New tools can be added or models swapped, without disrupting your workflow.


  • No Context Limits → Analysis is not restricted by token or context length, ensuring comprehensive results without truncation.


  • Zero Hallucination in Artifacts → Generated artifacts are guaranteed to be accurate and verifiable, eliminating risks of fabricated content.

Need Help?

If you have further questions, reach out at support@petavue.com

Did this answer your question? Thanks for the feedback There was a problem submitting your feedback. Please try again later.

Still need help? Contact Us Contact Us