This article is relevant if you are an executive sponsor, business leader, or NetSuite practitioner exploring MCP and want AI outcomes that are not only impressive, but more predictable, repeatable, and useful in the context of your specific NetSuite account.
TL;DR Summary
The MCP Standard Tools SuiteApp already provides meaningful operational value in NetSuite. Custom MCP tools take that value further, especially when the desired outcome depends on business-specific context, trusted retrieval patterns, and logic that should not be rediscovered every time a user asks a question. In our experience, this is where NetSuite AI becomes more portable, governable, and practical for broader adoption. The Prolecto Client Accomplishments AI Tool demonstrates how custom tools can help transform fuzzy historical work data into an executive-ready summary that supports strategic planning and relationship continuity across client engagements. The sample shown in this article uses the anonymized name Acme Inc. to protect client confidentiality.
Background
In my March 14, 2026 article, “Understanding MCP in the Context of NetSuite AI”, I noted that custom MCP tools deserved a closer look. This article is the promised follow-up. It extends that earlier discussion into a practical question: how should organizations think about custom MCP tools when they want more dependable outcomes?
This discussion also builds on Prolecto’s broader body of NetSuite AI work, including our earlier writing on connecting NetSuite to external AI platforms and our ongoing exploration of practical AI patterns inside the NetSuite ecosystem. The point is not simply that AI can be connected to NetSuite. The more important question is how to shape that connection so it produces results that business teams can trust and repeat.
At Prolecto, we are interested in more than getting a model to do something clever once. We care about recurring success; how to make AI-augmented work more useful to ordinary business users, more grounded in the realities of a client account, and easier to trust over time. That is the lens through which we view MCP custom tools.
Why NetSuite Custom MCP Tools Matter
At a high level, prompts communicate user intent in natural language. Agent skills and instructions help shape how the LLM approaches the work. MCP tools give the LLM secure ways to retrieve data and take action. The NetSuite MCP Standard Tools SuiteApp already provides real operational value because it exposes general capabilities that can help skilled users produce good outcomes. But the central challenge is not whether the model can do something useful once. The challenge is whether it can do so efficiently, predictably, and repeatedly across sessions and across users with different levels of expertise. General capability is not the same thing as account-specific understanding.
An LLM may know a great deal about business software and even something about NetSuite in general, but it does not inherently know the unique patterns, definitions, custom records, field conventions, and decision rules that make your account work the way it does. That gap matters. It shows up in the form of extra discovery by the LLM, extra calls to the NetSuite account which consumes integration bandwidth, extra tokens, and more variability from one session to the next. Once business users get past the euphoria of producing that meaningful report or triggering a successful record update in NetSuite using an external AI application, the question of how to do that across sessions and across users with different levels of expertise becomes top of mind. MCP Custom Tools offer a possible answer to this important question.
To recap, an MCP tool is a function that encapsulates logic to perform a specific operation such as creating a record, running a report, or doing something business-specific. In NetSuite, these tools are implemented using JavaScript just like other SuiteScript functions. The NetSuite MCP Standard Tools SuiteApp offers a set of predefined functions that AI Applications can invoke; custom tools are created by your team, a third-party vendor, or a NetSuite practice like ours to expose business-specific functionality to AI applications.
Key Benefits of Custom MCP Tools
MCP Custom Tools are a not-so-hidden gem that businesses can leverage to amplify their AI-augmented business outcomes. They have been available in NetSuite’s AI toolkit for a while but have not gotten as much attention as they deserve.
A custom tool allows a team to encode repeatable business logic once and expose it as a reusable function. The LLM can then focus on orchestration rather than rediscovery. That distinction is important:
“In many real-world scenarios, it is easier to steer the LLM towards invoking the right tools via skills and instructions than it is to rely on the model to reconstruct the same business logic every time.”
Here are some of the key benefits of using custom MCP tools.
- Context: They expose account-specific definitions, metadata, and business logic in a form the LLM can use more effectively.
- Efficiency: They reduce repeated schema exploration, query construction, and trial-and-error data retrieval.
- Predictability: They implement the business logic once so the LLM can focus on orchestration rather than rediscovering the same steps each time. While having a custom tool does not guarantee that the LLM will use it, if it does, the outputs are predictable and repeatable.
- Portability: They keep more of the business context on the NetSuite side, which makes it easier to work across different LLMs and AI platforms.
- Accessibility: They help average users get reasonable, repeatable outcomes without needing advanced prompting skills. AI solutions that require sophisticated superusers to produce reliable outcomes are generally not scalable.
- Business-first orientation: If designed properly, custom MCP tools enable solutions that are oriented around business workflows and user questions, not thin wrappers around low-level API endpoints.
Taken together, these benefits help move NetSuite AI from general capability toward dependable operational value.
Case Study: The Prolecto Client Accomplishments MCP Custom Tool
One of our goals as a firm is to help clients see and sustain the value of the work performed over time. Internally, our work is tracked through the Prolecto Task Manager (PTM), which wraps NetSuite case records to capture actionable work, stakeholder communication, and alignment. Timesheet entries summarize the work actually performed and also appear in client-facing billing. Together, these records hold a rich history of what was done and why it mattered.
Yet there is a difference between having historical records and having an executive-ready understanding of accomplishments. Leadership transitions happen. Strategic planning cycles happen. Sometimes it is essential to step back and answer a simple question: what meaningful accomplishments have we delivered for this client over the years by business function and by period? Before AI, we experimented with more human-centric approaches to maintaining accomplishment lists in custom record structures, but the cost of keeping them current was too high.
The Prolecto Client Accomplishments AI Tool was designed to solve that problem. The objective was to produce a concise, high-impact summary of meaningful accomplishments using PTM and timesheet data, grouped by year and business function, while excluding routine support, break-fix work, and low-value activity noise. The tool is not specific to one client; it is a reusable pattern we can apply across client engagements where historical accomplishment synthesis is valuable. Activities are not the same thing as accomplishments; only delivered outcomes, solved problems, implemented solutions, and strategic milestones belong in the summary.
We also liked this use case because it sits in the right middle ground for responsible AI design. It is fuzzy enough that human judgment matters, but controlled enough that business users can quickly assess whether the result is useful, grounded, and directionally accurate.
This is exactly the sort of Large Language Models, Fuzzy Problems, and NetSuite where custom MCP tools can shine. To a Prolecto expert, the underlying logic is understandable. We know what an accomplishment looks like in the context of our records, we know that a PTM is a NetSuite case, we know which data sources matter and how they are connected, we know which fields are trustworthy, etc. But asking an LLM to infer all of that from scratch each time, using only general tools, introduces unnecessary variability and cost (both in terms of speed and accuracy).
We created an MCP Custom Tool that exposes the relevant functions encapsulating the desired business logic and abstracts account-specific nuances from the LLM. Thus, instead of having the LLM explore the schema, guess the right query, and consistently interpret account-specific concepts, we hand it ready-to-use functions. The result is a cleaner path to the desired executive output. In the sample output below for the anonymized client “Acme Inc.”, I generated this report in no time in preparation for a call with the client who had been dormant for a little while and was seeking to re-engage. The final summary is grouped by year and business function and written in concise, action-oriented language suitable for executive review. This report helps us quickly get situated and focus on planning rather than relying on memory. It also offered the client an opportunity to react and possibly correct any misinterpretation. Cross-referencing the underlying PTMs facilitates human validation. The Prolecto Client Accomplishments AI Tool meets the business objectives it was designed to produce and is a very useful addition to our management suite.
Comparative Analysis
We also performed a comparative analysis of results produced with the MCP standard tools versus our custom tool. Using the same prompt in Claude, we first executed the request through an MCP connector with only the standard tools enabled. We then ran the same request through a second MCP connector with both the standard tools and our custom tool enabled. Our objective was to determine whether the custom tool delivered material benefits compared to relying on the standard tools alone.
The image above shows the summary-level outputs side by side. Notice that the custom tool execution found and analyzed more PTMs (71 versus 68) and significantly more timesheet entries (447 versus 216) than the standard-tools-only execution. The large variance in timesheet coverage reflects an important piece of historical context. In the early days of our PTM system, the linkage between PTMs and timesheets was optional and less tightly enforced. As a result, older timesheets, while still associated with a client, may not include a PTM linkage. The LLM was effectively blind to this nuance when using only general tools, which led to a substantial reduction in data reach. Arguably, this detail could be added to the prompt, but that would consume more of the model’s context window and still would not guarantee that the LLM would apply the knowledge correctly when constructing queries.
Unsurprisingly, the custom tool execution extracted more accomplishments because it had access to a richer and better-shaped dataset.
For additional comparison, we adjusted the prompt to request a few operating metrics, which are illustrated in the next image.
The custom tool approach not only performed better qualitatively; it also required fewer calls to NetSuite and completed the task significantly faster. Also, notice the tool invocation pattern. The standard-tools-only execution performed metadata discovery and ran multiple generic queries. By contrast, the custom-tool execution favored fit-for-purpose custom functions over generic retrieval functions. The result aligns with our thesis: when the business logic is encoded once in the right tool, the LLM can spend less effort rediscovering context and more effort producing useful outcomes.
MCP Custom Tools vs. Prompts, Skills, & Instructions
It is important to underscore that the use of MCP Custom Tools does not make other artifacts such as skills, instructions, and prompts less important. Quite the opposite; they are complementary. The shared objective of all these artifacts is to ground AI applications in relevant business context to produce desired outcomes as efficiently, predictably, and repeatedly as possible.
Prompts convey intent and should be as simple as the use case allows. Skills and project instructions provide guidance around tool selection, order of operations, and presentation. Custom MCP tools provide deterministic outputs rooted in business logic. The MCP Standard tools provide broad operational reach. Together, they create a layered model for repeatable outcomes: governed prompts, shared guidance, general operational capability, and business-specific execution.
NetSuite’s newer AI Connector Service Companion SuiteApp reinforces the same design direction. Oracle describes it as a centralized repository of enterprise-grade AI prompts designed specifically for NetSuite users for use with external AI applications. Users can copy prompts, open them directly in a supported AI application (currently Claude and ChatGPT), customize them in NetSuite, and save revised versions for reuse. This new SuiteApp solves for governed prompt management inside NetSuite.
Additionally, NetSuite has published an AI Connector Instructions Agent Skill as a way to add shared context and repeatable guidance so AI applications can interpret and run prompt samples more predictably. This solves for reusable behavioral guidance. We applaud these developments which align closely with our thesis:
“Stronger outcomes come from governed prompts, shared guidance, and reusable workflow patterns rather than leaving each user to reinvent the interaction from scratch.”
Custom MCP tools solve for the remaining layer: business-specific logic and context that should not be rediscovered from scratch every time the model runs.
It is worth noting that the Prolecto Client Accomplishments AI Tool was built before these more recent NetSuite AI offerings. Going forward, we intend to incorporate these newer NetSuite-managed prompt capabilities for an even better user experience; one that combines governed prompts in NetSuite, optional shared skills, and the existing business-specific custom tools in a more unified flow.
Best Practices for Getting Started with Custom MCP Tools
Custom Tools should not be created just because we can. They should be the result of deliberate solution design that aims at maximizing the value of the business’s AI investment in a specific fuzzy problem domain.
Besides getting hands-on experience with creating and deploying Custom MCP Tools, which is the subject of another discussion, this exercise exposed and reinforced a few insights.
- Always start with the outcome. Define the executive or operational result you want before thinking about technical implementation. In our accomplishments case study, the desired result was not “write a query”; it was “produce a useful historical accomplishment summary.” We had to refine and clarify our definition of an accomplishment and iterate until the outcome was satisfactory.
- Design for non-superusers. Broad adoption suffers when success depends on prompt craftsmanship. Reusable tools and skills can help normal users produce more reasonable outcomes with less variance.
- Build in a control mechanism. The best AI use cases give humans a straightforward way to assess output quality. The accomplishment summary is valuable in part because leadership can easily judge whether the result is thematically grounded and useful.
- Package reusable logic. Identify the parts of the work the LLM should not have to rediscover every time. Consider capturing repeatable sequences of operations or aspects that are critical to reliable results into custom tool functions. That often includes data retrieval paths, filtering logic, batching logic, and account-specific definitions.
- Prefer workflow-oriented tool design. The best custom MCP tools are designed around the questions users naturally ask and the workflows they actually perform. That is usually more valuable than exposing thin wrappers around low-level system functions.
- Let AI assist with the build. The LLM can help produce the custom tool itself. When you have produced a satisfactory outcome via thoughtful design or experimentation, put the LLM to work to produce a reusable custom MCP tool. That does not eliminate the need for judgment, but it can accelerate development once the business objectives and data model have been established.
Building Dependable NetSuite AI Outcomes
The larger lesson is that serious NetSuite AI work is not about treating the model like magic. It is about thoughtful modeling. The NetSuite MCP Standard Tools already create important operational possibilities. The Companion SuiteApp improves prompt governance and reuse. Skills add shared execution guidance. Custom MCP tools take the next step by encoding workflow-specific logic so that average users can get more consistent outcomes without expert prompting.
This is how organizations move from experimentation to operational advantage. It is also how they prepare for a world in which AI and agentic behavior continue to expand. Without guardrails and workflow-aware design, AI can amplify inconsistency just as easily as it can amplify productivity.
Our view is that the most valuable AI applications are not the ones that demand super users; they are the ones that help ordinary business stakeholders get reliable help from well-designed systems. That takes listening, modeling, technical competence, and execution discipline. That is also where Prolecto differentiates itself; not by producing isolated demos, but by operationalizing fuzzy business use cases responsibly and repeatedly.
Finally, this article has deliberately not focused on the mechanism and technicalities of creating MCP custom tools. For readers who want the hands-on version, we are considering a follow-up article with a step-by-step video walkthrough that shows how to create a NetSuite custom MCP tool from scratch, including the SuiteScript structure, the JSON schema, deployment considerations, and practical design choices. If that would be valuable to you, please let us know in the comments.
If you found this article relevant, feel free to sign up for notifications to new articles as we post them.
If you are ready to evaluate MCP opportunities in your NetSuite account and discuss how to model your own custom tools for repeatable outcomes, let’s have a conversation.
