In this recap article, we'll look at the strategies, lessons, and practical use cases that are helping enterprise sales teams get more value from MCP read + write today.
.png)
The Accord MCP Server lets AI tools like Claude, ChatGPT, and Cursor read and write inside Accord using plain language instructions, all while strictly adhering to the user's permissions. On a recent live session, Ryan Rich, Accord's Chief Customer Officer and Co-Founder, showed what that looks like in a live build.
Ryan connected Claude to Accord and built four workflows in one session: prepping content for a feature release, running a brand update across a content library, checking what's reported on deals against what's actually happening, and coaching a seller by updating a live Accord. This recap walks through each one, along with the permissions model and the tips that came up in the Q&A.
Model Context Protocol (MCP) is an open-source protocol that lets AI tools connect to other software without custom engineering. Anthropic created it and open-sourced it in late 2024. A tool like Accord publishes an MCP server, and any AI tool that supports MCP can connect to it and read and work inside that platform.
The Accord MCP Server launched as read-only in May and added write access in September 2026. Read access lets an AI tool query your Accord data. Write access lets it act on that data.
Accord is the execution layer for value-selling, so it helps to know the problem it exists to solve. "Most teams already have established sales methodologies and the gap is usually in enforcement," Ryan explained. "That's where inconsistency comes from." Accord helps teams set, enforce, execute, and score their methodology, and Accord Agents handle a lot of the busy work along the way. The MCP Server extends that to whichever AI tool your team already uses.
The session kicked off with permissions because they're the first question from anyone in RevOps or security. Ryan's answer was short: "Whatever your users can do in Accord, that's what MCP can do."
Accord has four standard roles: Workspace Admin, Playbook Admin, Content Admin, and Member. Enterprise customers can customize the permissions for each role, create their own roles, and set MCP access per user, all controlled by your workspace admin.
The design also protects against accidents. "If you give a sales rep MCP access, they're really not going to be able to do more than what they can do in Accord," Ryan said, so a seller can't change an admin setting by mistake.
Evan Prowse, Accord’s Product Marketing Manager added a practical way to start: create a role called something like "MCP Power User," give it to a handful of people, and see what's possible before you roll MCP out across your standard roles.
AI is good at looking across a large number of sources, checking whether information is current and on brand, and turning it into something people can act on. Ryan started with a scenario for anyone in enablement or product marketing: a new feature, Account Planning, is about to launch, and the team has a one-pager, a blog post, a knowledge base article, and a landing page.
He ran a skill that reads all of that content, works out what a seller needs to know, packages it in a consistent format in the Accord resource library, and publishes it to a product hub where sellers can find it and pull it into their deals.
When the run finished, the product hub had a new main page of enablement written by the AI, plus the customer-facing and internal materials: the PDF one-pager Ryan uploaded himself, links to the knowledge base and documentation, and the enablement overview. Every new release will keep landing in the same place in the same structure.
The second example was a rebrand. Ryan pretended the company was changing its name and asked the system to scan the entire library of customer-facing resources, list every inconsistency, and say how to fix each one. Once he approved the list, it replaced every out-of-brand mention with the new one.
At the end of a quarter, a sales leader may have 20 deals to review and no time to dig into each one. For this workflow, Ryan used a skill that combs through call recordings, emails, CRM data, and the content in each Accord. The source material can come from Gong, your CRM, or any other data lake.
The output was an interactive HTML dashboard covering the three biggest deals. For each deal it showed the account, amount, close date, and stage, along with a prediction of how likely the deal is to hit its expected close date and how accurate the reported information is.
The first deal showed why this matters. The system had high confidence in the close date, but not everything in the mutual action plan was up to date. A layer deeper, the net read was "likely to close," because legal and procurement were done and the package was about to go to the CFO. It also asked a pointed question: why isn't the budget holder in Accord? The CFO hadn't been added, which is a gap that's easy to miss in a CRM view. Ryan said sales leaders tend to get the most value from keeping this view high level, with the detail one click away.
This was the one Ryan called his favorite, and the reason Accord was built: giving AEs a better way to work with their customers, with AI as a partner that coaches them through the deal.
He set up a scenario. An AE has the biggest deal of the year, and the executive summary in their Accord is being shared with the executive team. After a few recent calls, the AE wants an independent check that everything is correct. Ryan asked Claude via MCP to review the Accord and flag anything that felt off.
Claude found that the summary section was out of date. All three numbers disagreed with what the customer confirmed on the previous day's call. It recommended an update, Ryan told it to apply the change, and it pushed the new summary live. After a refresh, the updated numbers appeared in the Accord. "This is exactly what our customer would see, now in their language, using their numbers," Ryan said.
The same check can run against any data in Accord, from resources to next steps, or anything connected to another system.
The first audience question was about skills: what's the advantage over typing exactly what you want each time? Ryan's answer is repeatability. A skill tells the AI what input to expect and what the output should look like, so a release prep that runs for every launch comes out the same every time. The seller updating one summary is a one-off, and probably doesn't need a skill.
If skills feel daunting, Evan pointed out you can tell Claude you want to build one and it will walk you through it. Ryan's approach is to do the task once, and if you like where it lands, ask Claude or ChatGPT to turn it into a reusable skill.
The Accord MCP Server provides read and write access to Accord data from within AI tools such as Claude, ChatGPT, and Cursor. It uses the Model Context Protocol standard, so no custom engineering is required to connect.
Any AI tool that supports MCP can connect. Ryan used Claude in the session and named ChatGPT and Cursor as other examples.
Reach out to our team, or directly to your CSM if you're already a customer of Accord, and we'll share Ryan's prompts and skills.
The through line across all four workflows is control and consistency. MCP ensures your AI works seamlessly with your Accord data and deals, strictly bound by your sales methodology and in-app permissions. With MCP, the work most teams know they should be doing, like keeping resources current, checking deals against reality, and catching what's out of date before a customer does, gets done consistently instead of only when there's time.
To go further, watch the full recording, or learn more about the Accord MCP Server. If you'd like to see how it would work for your use case, book a demo with our team.
.png)
.png)
.png)