Using Ollive
Access and security.
Your login. Your workspace.
Connecting an app uses your Ollive login. There are no keys to create, and the app sees only the workspace you choose.
How signing in works
Apps connect to Ollive with OAuth, the standard way to let an app act for you. You sign in on Ollive’s own page, so the app never sees your password and there’s no key to hand over.
- Your app opens Ollive
Ollive opens in your browser. Sign in the way you always do.
- Ollive asks you
The page shows which app is asking, what it will be able to do, and which workspace it will use. If you belong to more than one workspace, choose one.
- You decide
Select Allow and you’re sent back to your app, connected. Select Cancel and nothing is shared.
What a connected app can do
| Read | Your workspace’s monitoring results: agents, conversations, verdicts, violations, metrics and transcripts. |
|---|---|
| Send | Conversations you give it, to be checked against an agent’s rules. |
| Acts as | You. It sees what you can see in Ollive, in the workspace you chose. |
| Can’t | Change your agents, rules, integrations or settings, or see any other workspace. |
Workspaces and teams
Each connection is for one workspace. To use another, sign out of Ollive in your app and connect again. Each app guide shows how.
Anyone in your workspace can connect their own apps. Teammates can share a setup file, such as a project’s .mcp.json, and each person still signs in with their own login.
Staying connected
Your app keeps itself signed in while you use it. Every so often, it asks you to sign in to Ollive again, which takes one click.
To disconnect an app, remove Ollive from it, or sign out of Ollive in it. To cut off an app’s access straight away, for example on a lost laptop, contact support.
Conversation text is data
Transcripts and evidence are your end users’ and agents’ own words. Ollive tells your assistant to treat that text as information, never as instructions to follow, and its tools send a conversation only when you ask.
- Only connect apps you trust.
- Read what your app asks you to approve, especially before it sends conversations.
- If you let your app run tools without asking, keep that to the reading tools.
Questions
- Do I need an API key?
- No. You sign in with OAuth, using your Ollive login. API keys are for the Ollive API.
- Does Ollive see my chats with my assistant?
- No. Ollive receives only the requests your assistant makes with its tools, such as “list the violations for this agent”, and any conversations you ask it to send.
- Do conversations I send count toward my agents’ results?
- No. They’re labelled MCP and kept apart from the conversations Ollive collects from your agents, so tests never mix with production.
- Is it the same data as the API?
- Yes. Each tool returns the same results as the matching API endpoint.