You can now run Orisu from ChatGPT
Orisu's MCP connector now works in ChatGPT. Add one URL, sign in, and build and run your content workflows without leaving the chat.
Orisu already spoke MCP, the open standard that lets AI assistants plug into real tools. This week that connection started working from ChatGPT too. Add one URL as a custom connector, sign in with your Orisu account, and ChatGPT can search your agents, build new workflows from a description, trigger runs, and show you the live canvas, all inside the chat.
If you tried this before and it failed right after sign-in, that was our bug, and it's fixed. There's a short note on it at the end.
What shipped
ChatGPT can now connect to Orisu over MCP. In practice, the assistant you may already spend half your day in can:
- Find and run your existing workflows: "Run my product-shot agent on this photo."
- Build new ones from a plain-language brief: "Make a workflow that turns a blog post into three social captions and a header image."
- Check on runs and recover from failures: "That run failed, what happened?" gets a real answer, because the connector exposes the typed error and the recovery options.
- Render the canvas in chat, with live per-node status, so you can watch a run move without switching tabs.
Under the hood this is the same MCP server that already serves Claude. Anthropic designed MCP as "an open standard for connecting AI assistants to the systems where data lives", and this is where that design pays off for us. We built the connection once, and ChatGPT can use it too.
Why we wanted this
Most marketing teams don't live in one AI assistant. Copy gets drafted in ChatGPT, research happens in Claude, and production happens somewhere else again. Moving work between them means copying and pasting, and some context gets lost on every trip.
With your production canvas connected to the chat you already have open, you skip that trip: the brief, your feedback, and the run sit in one thread. ChatGPT's own MCP support is built for this. OpenAI's connector documentation describes remote MCP servers as a way to "give models new capabilities" to connect to and control external services, with approvals required before data is shared. Write actions ask for confirmation by default, so the assistant proposes and you approve before anything runs or spends credits.
How to connect
It takes three steps and about two minutes:
- Turn on developer mode. In ChatGPT on the web, open Settings → Security and login and enable developer mode. Per OpenAI's guide, it's available on Pro, Plus, Business, Enterprise, and Education accounts, and it's what gives ChatGPT "full Model Context Protocol (MCP) client support."
- Add the connector. Create a new connector for a remote MCP server and paste the Orisu endpoint:
https://api.orisu.ai/mcp. Choose OAuth for authentication. - Sign in and try it. A sign-in window opens; use your normal Orisu account and approve the scopes. Then ask something small: "List my recent Orisu runs." When the answer comes back with your actual runs, you're connected.
The MCP quickstart has the full walkthrough for both ChatGPT and Claude, including what to ask first.
If an earlier attempt failed
Some early connection attempts failed right after the sign-in step. The cause was on our side: our OAuth discovery metadata drifted from the spec in ways that strict clients like ChatGPT reject. That metadata is now spec-correct. If you hit the failure, remove the Orisu connector in ChatGPT and add it again so it refetches the fixed metadata; a stale connector keeps the old copy.
Full notes are in the changelog. If you're new to workflows themselves, start with how to build an AI content workflow, then come back and try building one from chat. Everything the connector creates lands on the same canvas your team already uses, so a workflow you build in a conversation doesn't stay stuck in that chat thread.
See the whole workflow.
Every step on Orisu is a node you can see, rewire and rerun. Templates are real share pages — open one and inspect the graph.
Common questions.
Do I need a paid ChatGPT plan to connect Orisu?
Yes. Custom MCP connectors in ChatGPT require developer mode, which OpenAI makes available on Pro, Plus, Business, Enterprise, and Education accounts on the web. Free accounts can't add custom connectors yet.
I tried connecting ChatGPT to Orisu before and it failed. What changed?
Earlier attempts could fail right after sign-in because our OAuth discovery metadata strayed from the spec in ways ChatGPT rejects. The metadata is now spec-correct for strict clients. Remove the old connector in ChatGPT and add it again so it refetches the metadata.
Can ChatGPT spend my Orisu credits without asking?
Not by default. ChatGPT treats tool calls that change things as write actions and asks you to confirm each one before it runs. You review the exact call, including which workflow and which inputs, before any credits move.
Does this replace the Orisu studio?
No. The connector renders a live canvas in chat and handles building, running, and checking on workflows. For fine-grained editing, node-by-node config, and reviewing outputs side by side, the studio is still the better place.
