Your workspace, inside the AI assistant you already use
Connect Claude, ChatGPT or any MCP client to Octaviaflow with one URL. Ask what ran, why it failed and what is connected — and get the answer from your real workspace, not a guess. Then tell it to run the thing again.
Ask in the assistant you already use
Add one URL to Claude, ChatGPT or any MCP client and ask in plain language: "did the Salesforce sync run last night?", "why did that flow fail?", "what are we connected to?". The assistant reads your actual workspace and answers from it — no dashboard, no tab switching, no copying ids between windows.
Your projects, flows and runs
List projects and flows, open a flow to see its steps and current state, page through recent runs, and pull up a single run to find the step that failed. The same view your team has in Octaviaflow, available to the assistant that is already helping you write the follow-up message.
Connections and agents, without the secrets
See which third-party apps a workspace has connected and whether each is healthy, plus the AI agents and knowledge bases configured in it. Credentials are never returned — not redacted, not obscured, never sent. An assistant can tell you a connection is failing; it cannot read the key behind it.
Plan and usage, before you commit
Ask what the workspace plan allows and how much of it has been used this period, so the assistant can tell you whether the thing you are about to build fits — rather than finding out at the limit.
And it can act, not only look
Run a flow now, stop one that is going wrong, pause an automation that is firing when it should not, or turn on a draft you have finished building — all from the assistant, in the same conversation where you worked out that it needed doing. Every action runs as you, under the same permissions and plan limits as the app, and your AI app asks you before it does anything that changes something.
Signed in as you, scoped to your workspace
Connecting opens the normal Octaviaflow sign-in and asks you to approve. The assistant receives a short-lived token for one workspace, and the same permission rules the app enforces apply to every request. You can disconnect an app at any time from Settings.
Built to the current spec
A remote MCP server over Streamable HTTP with OAuth 2.1, PKCE and Client ID Metadata Documents — the 2026 specification, not an SSE relic. Any compliant client can connect; nothing is specific to one vendor.
Frequently asked questions
What is MCP?
The Model Context Protocol is an open standard that lets an AI assistant use an external service. Octaviaflow runs a remote MCP server, so an assistant like Claude or ChatGPT can read your workspace and answer questions about it.
How do I connect it?
Open Settings → MCP Access in Octaviaflow, copy your workspace endpoint, and add it as a custom connector in your AI app. You sign in through Octaviaflow and approve the connection — it takes about a minute, and the page walks you through each app.
Do I need to paste an API key?
No. Connecting uses OAuth 2.1, the same way you would connect any other app. The assistant receives a short-lived token it refreshes on its own, and you can revoke it from Settings without changing anything else.
Can the AI see my connector passwords or API keys?
No. Credentials are never returned over MCP. An assistant can see that a connection exists and whether it is healthy; it cannot read the secret behind it.
Can the AI change or run things?
Yes, within limits. It can run a flow, cancel a run, and pause, resume or activate a flow — nothing else. It cannot edit a flow, delete anything, change billing, or touch your connected credentials. Each action runs as you, under the same permissions and plan rules as the app, and your AI app confirms it with you first.
Which plans include it?
Every plan, and the 14-day free trial. Reading your workspace over MCP does not consume credits.
Which AI apps work with it?
Any client that supports remote MCP servers over Streamable HTTP with OAuth — Claude and ChatGPT today, and others as they add support. Nothing about the endpoint is vendor-specific.
Ask your assistant what your workflows are doing
Create a workspace, connect your AI app, and ask it a question about your own data.