RunEvr
Agentic project management environment for creatives
Start new thread

Your AI is about to get a RunEvr login 🔑

by•

We're genuinely excited about this one.

You already start your day inside an AI - Claude, ChatGPT, Cursor. That's where the thinking happens. But the work lives somewhere else, and you spend all day copying between the two.

So we're killing that gap: RunEvr is getting an MCP server. (MCP = the open standard - now backed by Anthropic, OpenAI, Google & the Linux Foundation - that lets your AI securely plug into real apps. RunEvr is becoming one.)

Connect the AI you already use, and just talk:

"Create a task for Sara to review the trailer, due Friday."
"Did anyone reply to my thread about the poster?"
"What's waiting on me right now?"

Your AI does it, and it shows up live in RunEvr for the whole team.
Works with Claude, Cursor, and ChatGPT.

The best part: you don't pay us a penny extra. It runs on your own AI account - your existing free or paid Claude/ChatGPT plan. We don't meter it, don't upsell it, don't touch your bill.

And one line we won't cross: your AI can never see more than you can. Same permissions as you - superpowers, not a backstage pass.

Which AI do you live in - and what's the first thing you'd have it do in your workspace?

---------------------

👇 If you’re just getting familiar with MCP, here are a few places to start:



14 views

Add a comment

Replies

Best

Answering your question first: Claude Code, most of the day. The first thing I'd ask is what's waiting on me right now, because read-only is the part you trust on day one.

I've built an MCP server, so a couple of notes from that side of it.

Your permissions line is the right principle and I'd push it one step further. Same permissions as the user is necessary but not sufficient, because a person holding forty permissions typically exercises about five of them. An agent will use all forty, cheerfully, because nothing in a tool description marks which ones are socially unusual. Scoping stops it reaching other people's data. It doesn't stop it doing far more, far faster, than that human ever would with their own access.

Which matters most on the write side. Create a task for Sara due Friday is the example that isn't like the others, because it generates work for a second person who wasn't in the conversation and has no idea an AI produced it. Worth marking agent-created items as agent-created in the UI. Cheap to build, and it lets Sara weigh that Friday properly rather than assuming a colleague chose it.

One practical thing that will otherwise cost you support tickets: in Claude, a newly connected MCP server's tools only load at session start. People will connect it, not see the tools, and tell you it's broken. Put restart your client in the setup flow and you'll save yourself a week of that.

 Dale - thank you, this is genuinely helpful.

You're right, and the sharp version of it is on writes. A person could do 40 things and does maybe 5; an agent will do all 40 - and through MCP it acts as you, so "create a task for Sara" just looks like you did it. So we're going to mark agent-created items as agent-created, so people can see it.

And restart-your-client is going into the setup flow - that one would've cost us a week.

Really appreciate you taking the time. 🙏

  Your version is sharper than mine, and it's the part that matters: it acts as you, so at the point of receipt there's nothing to distinguish it.

One thing on the labelling, since you're building it anyway. It only works if it survives to where people actually read the task. Most people don't meet a new item in the app first, they meet it in the notification, the email, or the daily digest. If the agent-created marker only lives in the web UI, Sara still gets a message that reads exactly like a colleague set her a Friday deadline, and the assumption forms right there, before she ever opens RunEvr. Worth carrying it into the notification surface as well as the item itself.

And I'd have the label name the person rather than the technology. Created by Dale's assistant tells Sara who to go back to. Created by AI doesn't, and she's still left working out whose deadline it actually is.