TaskRay MCP, now in beta, lets users ask their AI assistant to take action in TaskRay using plain English, whether that means checking project status, finding assigned work, creating a project from a template, logging time, or prepping for a customer call. It works inside the user’s org, permissions, and live records, giving teams a faster way to access and manage project work without leaving their preferred AI assistant.
Think about last Monday morning. Before you could actually start delivering for a customer, you probably lost 20 minutes just figuring out where things stood: clicking into a project, scanning task lists, and piecing together what was blocked across three different tabs. Then doing it again for the next project.
None of that is the job. The job is delivering for the customer.
Here’s the thing: TaskRay already solved the hard part. The plan, the milestones, the dependencies, the status; it’s all structured and current, not scattered. What’s changed is where people expect to reach that structure from. You’re already talking to an AI assistant in your browser, in Slack, or on your desktop about everything else. It makes sense that TaskRay should be one of the systems it can reach into as well, alongside the app you already know.
TaskRay MCP is now in beta. Ask your AI assistant for what you need from TaskRay in plain English, and it goes and does it, right in your org, as you, with your permissions, on your live records. No new app to learn. No export to clean up. You just ask.
Here’s what that looks like:
“What’s the status on Acme?” A plain-language rundown of what’s done, what’s blocked, and what’s due next. No clicking into the project at all.
“What do I own this week?” Everything across every project you’re on, ordered by deadline, with blockers called out. Your whole Monday, answered before your coffee’s done.
“Set up a new project for Acme using the Enterprise Onboarding template, starting the 3rd.” Your assistant builds the full project structure and schedules the plan using the same templates and logic TaskRay already runs on. You skip the screens and get straight to a plan that’s ready to go.
“Here’s what I worked on this week. Log my time.” Describe your week the way you’d tell a coworker. Your assistant maps it to the right tasks, drafts the entries, and waits for your approval before filing anything.
“Get me ready for my call with Acme.” Project status, recent email threads, and notes from the last call, all pulled into one brief before you dial in.
TaskRay is still where your plans live and where the deep work happens. Now it’s also one prompt away from wherever you’re already working.
Salesforce manages the relationship with your customer: the deal, the account, the contract. The moment that deal closes, a different question takes over: what did you promise them, and are you keeping that promise? That’s a delivery question, and it’s the one TaskRay has always answered.
We’re leaning further into that role. TaskRay is becoming the AI-ready execution system of record for everything that happens after the sale—the structured source of truth an AI assistant can actually trust when it needs to know what’s owed, what’s blocked, and what’s next.
That structure is the whole point. Point an assistant at raw data, and it has to guess what a “project” even means or how a milestone relates to a dependency. Point it at TaskRay, and it inherits a model battle tested across more than a thousand implementations. It doesn’t guess. It knows. That’s what TaskRay MCP opens up: not just access, but understanding.
TaskRay MCP runs as you, the signed-in user. Every sharing rule and field-level permission that applies to you in the app applies here, too, automatically. Your assistant sees what you’re allowed to see. Nothing more.
Every action it takes is attributed back to you. Whether you updated a task yourself or asked your assistant to do it, the audit trail reads the same: same TaskRay, same rules, just a new way of asking.
We’re not betting on one AI platform winning, so TaskRay MCP doesn’t either. Connect Claude, ChatGPT, Copilot, or Agentforce—whatever your company has already standardized on. Same connection, same capabilities, your choice.
If you’ve got the resources to build, MCP is the front door: connect it once and anyone on your team can just ask, no code required. And this is only the start of what’s possible on this foundation.
TaskRay MCP is in beta because we want real teams in real orgs with real customizations putting it to work before we open it up wide. Your feedback shapes what ships next, so getting in early is the best way to make sure this works the way your team actually needs it to.
If you want in on the first group, register here for details and next steps.
