Brownsmith Dynamics

Services, products, company information, learning, and contact paths in one place.

Why a service based business needs an MCP

A service business can give its AI assistant better working knowledge with a small plugin, then add connected data and actions only when the work calls for them.

October 2, 20264 min read
Business documents and a calculator arranged on a desk

Product perspective

Workflow Automation Hub

View product

A service business sells judgement, but a surprising amount of the work around that judgement is repeatable: explaining the offer, qualifying an enquiry, preparing for a call, and keeping follow-up from slipping. An AI assistant can help with those jobs only when it understands how this particular business works.

The title says MCP because that is where connected business tools enter the picture. Our first package is simpler: a plugin made from skills. It gives an assistant reusable instructions and reference material. MCP becomes useful later, when the job needs current records or approved actions in a business system.

The Service Business Case

Good service depends on context that is easy to lose.

A consultant, agency, accountant, or specialist trades firm does not need an assistant that can write a generic email. It needs one that knows the offer, ideal client, and what to check before making a promise. Without that context, every prompt starts from scratch.

I wanted this toolkit to reflect the work Brownsmith Dynamics is moving towards: Business Efficiency and Organic Lead Generation. If someone brings us a tangled workflow or an enquiry problem, the assistant should start with that business and its constraints. It should ask where work gets stuck, separate facts from assumptions, and choose a useful first change. Jumping straight to an AI tool misses the point.

A small plugin puts the working method in one place and cuts repeated briefing. The team can refine it as it learns what helps. A person remains responsible for the advice and final communication.

The Current Package

This first version carries instructions, not live connections.

The four skills are business-diagnosis, workflow-improvement, organic-growth-audit, and implementation-brief. They move from understanding the problem to improving a recurring job, reviewing the organic enquiry path, and scoping delivery. Three starter prompts point to the first three; the brief can be requested when a business is ready to scope a change. Shared references carry our approach and a quiet contact footer, so those details stay consistent across the skills.

The portable package has one root plugin.json, a skills/<name>/SKILL.md file for each skill, shared references, and a README. The manifest identifies the plugin; the skill files explain when and how to apply each workflow. This version needs a compatible AI host, the packaged instructions, and context the user chooses to supply. It needs no hosted MCP server, API key, or account connection, and has no access to a CRM, inbox, calendar, or customer records through the package.

To distribute it, package that folder as one ZIP with the manifest, skills, and shared references together. Test complete and incomplete business requests before publishing. For ChatGPT submission, use the developer portal with an eligible verified identity, listing information, an icon, and the relevant support and policy links. Uploading creates a draft; validation and the applicable publication checks still follow. Our source is available on GitHub, which does not by itself mean the plugin has been approved or listed.

I also wanted a dedicated plugin page where someone could understand the four skills and find support, privacy information, and terms before using the package. For our listing, that page belongs in extensions.com.openai.interface.websiteURL. The source repository and the publisher's general homepage serve different purposes; setting homepage or author.url does not populate websiteURL. OpenAI requires all four listing URLs for MCP review, while skills-only metadata validation does not require all four. We provide them because they make the toolkit easier to understand and assess.

That limit is deliberate. Written guidance can help with a draft or a planning task without being trusted to change business data. The owner remains responsible for keeping prices, service descriptions, policies, and examples accurate. If a fact has not been approved, the assistant should treat it as an assumption and ask for review rather than turn it into a confident claim.

A Later Connection Layer

Add MCP when the job needs current information or an action.

A separately packaged connected version could expose read-only knowledge or status first: perhaps approved service information, a project status, or whether an enquiry has been assigned. A later, carefully scoped workflow might prepare a follow-up for approval or create a task after a person confirms it. Those are possible designs, not capabilities in this release. OpenAI currently requires an MCP server to be included in the initial submission of an MCP-backed plugin; it cannot simply be added to an existing skills-only listing.

MCP is useful when an assistant needs a defined interface to tools or data. It does not replace the plugin's working instructions, and it does not make access safe by itself. The business still needs to choose which records are visible, who may request an action, and where a person must approve it. An optional app interface may help when staff need to inspect a queue or review changes; it adds little when the assistant only needs guidance.

The format depends on the friction: a skills-only plugin for repeatable guidance, a read-only MCP for current facts, or a controlled action workflow that crosses into business systems. A local firm might start with enquiry preparation; a consultancy with project status; an agency with a reviewed lead handoff. Those are possibilities, not features of this release.

Conclusion

Start with the knowledge the business can stand behind.

A service business benefits from an AI plugin when it captures useful working knowledge and makes routine work easier to start. That is the job of this first toolkit. If the next bottleneck is stale information or a manual system update, then consider a narrow MCP connection with clear permissions and human review.

The sequence matters: agree on what is true, teach the assistant how the work should be approached, and connect systems only when there is a defined reason. The person who owns the service still owns the knowledge and the decision.

Public Toolkit

Brownsmith Dynamics Business Improvement Toolkit

Explore the four skills, example requests, package limits, and support and policy links.

Explore the plugin

A Connected Interface

Plan a useful MCP boundary

See how Brownsmith Dynamics approaches a narrow, reviewable MCP for an existing application.

Read our MCP approach

Help and Policies

Support for the business improvement toolkit

Find help with the four skills and links to the plugin's privacy policy and terms.

Visit plugin support

Research notes

Sources and Supporting Material

These references support factual claims in the article. Brownsmith's interpretation and forward-looking analysis remain editorial judgement rather than vendor promises.