Skip to main content
Your applications are connected. What’s left is pointing your AI agent at the gateway. After that, nothing about how you work changes - you ask, the agent acts, and Oleria applies your permissions and your company’s policy to every call.
Connect your applications first. An agent pointed at the gateway can only reach applications you’ve already authorized. See User setup.

​
Add the gateway to your AI agent

The gateway is exposed to agents as an MCP connector. Any agent that supports remote MCP connectors can use it - including Claude and Claude Code.
1

Pick the agent you'll use

Any agent that supports remote MCP connectors works. These steps use two as the examples: Claude in a browser at claude.ai, and Claude Code in a terminal. Where the two differ, both are shown.The sequence is the same for any other MCP-capable agent - only where you add the connector changes.
2

Look for the gateway in your connectors

Open your agent’s connector settings. In Claude, this is Settings -> Connectors.If your administrator has already published the gateway, it’s listed with an authorization of Individual - meaning you still sign in as yourself. Skip to the next step.Claude connector settings listing the Oleria AI Gateway as a custom web connector with individual authorization
3

Add it yourself if it isn't listed

You don’t need administrator access to add the gateway. Select Add, choose a custom connector, and give it a name - this is what you’ll see in your connectors list - along with your gateway’s MCP endpoint:
In Claude Code, add it from the command line instead:
--scope user makes the gateway available in every project you work in; leave it out to add it to the current project only. Run claude mcp list to confirm it registered.This endpoint is the same for every Oleria production tenant - there is nothing tenant-specific to substitute.This is the same endpoint an administrator publishes centrally, so a connector you add yourself behaves identically: it signs you in as yourself, reaches only the applications you’ve connected, and is governed by the same policy.
4

Sign in to Oleria

Select Connect. In Claude Code, run /mcp and choose the gateway instead. Either way, the sign-in opens in your browser.Enter your Oleria workspace URL and select Continue.The Oleria sign-in screen asking for your workspace URL, with a Continue buttonThen sign in with your identity provider, exactly as you would to reach Oleria normally. There are no passwords, tokens, or API keys to paste into the agent.
5

Activate the agent

Signing in brings you to the Agents tab of the AI Agent Gateway in the Governance App, where your agent is listed as Waiting to be activated. Select the activation button - it carries the name the agent registered under.The Agents tab in the Governance App showing an agent waiting to be activated, with an Activate button, above an already connected agentYour browser then returns to claude.ai, or to Claude Code in your terminal, and the connector shows as connected.Activation is what ties the agent to you. From this point the gateway acts as you, with your permissions and the applications you connected above. The agent stays listed on this tab with the address it last called from, when it last ran, and when it connected - and that is where you go to check what is connected in your name.
6

Review the tools

The connector exposes one tool per supported application - github_api, jira_api, slack_api and so on - plus the helpers list_apps, get_connection_info and get_credentials. Choose whether each tool runs automatically or needs your approval first.An application you haven’t connected still appears in this list. Listing a tool isn’t access - calling one for an unconnected application returns an error telling you to connect it first.Connector settings showing the Oleria AI Gateway connected, its MCP URL, and tool permissions for Confluence, Github, Google, Jira, and other tools

​
Work with your agent

Once the connector is in place, nothing about how you work changes. Ask for what you need, and the agent uses the gateway to reach each application in turn. Agent session connected to the Oleria AI Gateway, running a request to draft a pipeline update from Salesforce opportunities and notes in Drive and save it to SharePoint A single request often spans several applications. Above, the agent reads open opportunities from Salesforce, gathers the related meeting notes from Google Drive, and saves the draft to a SharePoint site - three applications, three sets of your permissions, and no credential in the agent at any point.

​
When an action needs approval

If your administrator marked an action as Ask, the agent can’t complete it on its own. Reading from Salesforce and Drive above is fine, but saving the draft is a Write file action on SharePoint. The gateway holds that call, records an approval request against the named approver, and tells the agent what happened. Agent session reporting that the Write file action requires human approval and that an approval request has been sent to the manager If the approver responds within about 30 seconds, the original call finishes on its own and the agent continues as though nothing had paused. Past that, the gateway returns a pending result, and you ask the agent to retry once the request is approved. A denied request never runs.
The gateway does not send the approver an email, a Slack message, or a Teams message. How your approvers are told depends on how your organization has set the gateway up, so agree that with your administrator before relying on Ask for anything time-sensitive.
Actions marked Deny are refused outright, and the agent is told the action isn’t permitted rather than being left to retry.

​
See what your agent did

The Activity tab shows every call your agents made through the gateway - the time, the agent, the session, the application, and the exact action. Connections and disconnections appear alongside API calls, so you can see how access was obtained as well as how it was used. Activity tab listing gateway calls with time, agent, session, app, and action columns Because the agent acts as you, every entry ties back to you and the session it ran in - not to a shared token.

​
Contact us

For questions, contact us at support@oleria.com.