Guides
Guides

Configure

What your institution needs in place, and how to add the connector to your AI assistant.

Endpoint: https://mcp.ncino.com/mcp

Setup at a glance

Setup has three parts. None of them is complicated, and institutions have typically completed all three in about 30 minutes.

Steps 1 and 3 involve nCino. You can work through them asynchronously, but we recommend doing them on a call with your nCino team so anything unexpected gets sorted out on the spot.

Start by contacting your Customer Success Manager, who will set this up with the right teams.

StepWho drives it
1Connect your identity providerYour IdP or IT administrator
2Add the connector in your AI assistantYour nCino administrator(s)
3Connect your nCino environmentYour nCino administrator(s)

You don't have to wait for step 3 to get started. Most of the release, schema, and documentation tools read nCino's own catalogs, so they work as soon as your assistant is connected.

Step 1: Connect your identity provider

Your administrators sign in on your identity provider's own hosted sign-in page, with their normal corporate credentials. To make that work, your identity provider is federated to nCino's identity platform once, up front. Your IdP administrator creates an application registration for nCino, releases a small set of standard user attributes, and registers a callback URL that nCino provides.

nCino has onboarding guides for the common identity providers, covering both OIDC and SAML 2.0. Your onboarding contact will point you at the right one.

If your institution already uses SSO with nCino, much of this may already be in place. Confirm with your onboarding contact before starting fresh.

The one thing that most often trips this up is a mismatched callback URL. It has to match exactly, including protocol, domain, and any trailing characters.

Step 2: Add the connector in your AI assistant

You sign in through the nCino Identity Platform in a browser and approve access. Sign-in lasts for a working period and then prompts again. Some assistants handle that more gracefully than others. See Limitations.

Microsoft Copilot Studio

Prerequisite: your agent must have generative orchestration turned on. MCP tools will not work without it. Your Power Platform data policies also apply, so if your administrators restrict connectors in the Power Platform Admin Center, those restrictions govern this connection too.

Build the agent in the classic experience. Agents built in Copilot Studio's new experience run on the GitHub Copilot harness, which bills Copilot Credits for building and testing as well as for running. In an environment without credits, setup stops with "You need credits to continue." To switch, turn off New experience on the Copilot Studio home page. An agent built in the new experience can't be moved to the classic one.

  1. Go to the Tools page for your agent and select Add a tool, then New tool, then Model Context Protocol.
  2. Fill in Server name, Server description, and Server URL. Use https://mcp.ncino.com/mcp as the server URL.
  3. For authentication, select OAuth 2.0, then Dynamic discovery. The nCino MCP supports discovery, so Copilot Studio finds what it needs and registers itself. You do not need a client ID or client secret.
  4. Select Create, then Next.
  5. On Add tool, select Create a new connection, then Add to agent.

Keep each person on their own sign-in. In the tool's Details, under Additional details, keep Authentication set to End user, which is the default. Everyone who uses a shared agent then signs in to nCino as themselves. If it's switched to Maker-provided, everyone who can chat with the agent works through the maker's nCino sign-in. Power Platform administrators can enforce end-user authentication for an environment. Per-user sign-in works in Teams, SharePoint, and custom websites, but not in the demo website.

See Microsoft's Connect your agent to an existing MCP server and Switch between classic and new agent experiences.

Claude

On a Team or Enterprise plan, an Owner adds the connector once and members then connect to it individually.

Owner:

  1. Go to Organization settings > Connectors and select Add.
  2. Hover over Custom, then select Web.
  3. Enter a name, such as nCino, and the server URL https://mcp.ncino.com/mcp. Select Continue.
  4. Keep the authentication settings Claude detects and select Continue. Don't choose No sign in. If you're asked to choose an OAuth client, choose Register automatically. You don't need an OAuth client ID or secret.
  5. Select Add.

Each member:

  1. Go to Customize > Connectors and find the connector in the list. It usually has a Custom label.
  2. Select Connect and complete the browser sign-in.
  3. In a conversation, select the + button, then Connectors, and turn on nCino.

On a Pro or Max plan, go to Customize > Connectors, select + Add, then Add custom connector, and follow the same steps.

Connectors cannot be edited after they are added. To change anything, remove it and add it again.

See Anthropic's Get started with custom connectors using remote MCP.

ChatGPT

Custom MCP apps require developer mode. Full support, including write actions, is in beta on Business, Enterprise, and Edu plans, on the web only.

  1. A workspace admin turns on developer mode in Workspace settings > Permissions & Roles, under Connected Data. On Enterprise and Edu, the person creating the app then turns it on for their own account in Settings > Apps > Advanced settings. On Business, only admins and owners can use developer mode, and each one turns it on for themselves. If a step is missed, the app will not appear.
  2. In Workspace settings, or in your own Settings if you're an enabled member, go to Apps, then Create.
  3. Provide https://mcp.ncino.com/mcp as the endpoint, along with a name, and select the OAuth authentication option.
  4. Select Scan Tools and complete the sign-in when it appears.
  5. Select Create. The app appears as a draft.
  6. An admin or owner publishes it from Workspace settings > Apps > Drafts. On Enterprise and Edu, Configure Actions and Configure Access let you choose which actions are allowed and which groups get the app.

ChatGPT keeps a snapshot of the connector's tools from when the app was published. If calls start failing after an nCino update, an admin can refresh the app's actions in Workspace settings > Apps.

See OpenAI's Developer mode and MCP apps in ChatGPT.

Gemini CLI

Gemini CLI now needs a Gemini Code Assist Standard or Enterprise license, or a paid Gemini API key. Since June 18, 2026, Google no longer serves it to free, Google AI Pro, or Ultra users, and points them to Antigravity CLI.

  1. Add the connector:
   gemini mcp add --transport http ncino https://mcp.ncino.com/mcp

Add -s user to configure it for your user account rather than the current project. You can also add it by hand to ~/.gemini/settings.json:

   {
     "mcpServers": {
       "ncino": {
         "httpUrl": "https://mcp.ncino.com/mcp"
       }
     }
   }

No other settings are needed. Gemini CLI works out how to sign in on its own.

  1. Start Gemini CLI. The first connection opens a browser for you to sign in. You can also trigger it with /mcp auth ncino.
  2. Run /mcp to confirm the server shows as CONNECTED and to see the available tools.

Gemini CLI needs to open a browser on your machine and receive the sign-in response on http://localhost:7777/oauth/callback. It will not work on a machine with no browser, over SSH without display forwarding, or inside a container without browser access.

See Google's MCP servers with the Gemini CLI.

Perplexity

  1. Go to Account settings > Connectors. To add it for your whole organization instead, an admin goes to Enterprise settings > Permissions > Connectors permissions.
  2. Select + Custom connector, then Remote.
  3. Enter a name, such as nCino, and https://mcp.ncino.com/mcp as the MCP Server URL.
  4. Choose OAuth for authentication and Streamable HTTP for transport. Leave the client ID and secret blank.
  5. Check the acknowledgement box and select Add.
  6. Select the connector card to sign in and turn it on.

On Enterprise, members can add their own custom connectors only after an admin turns on Allow members to add custom connectors, which is off by default. An admin can instead add the connector for the organization and share it from Enterprise settings > Permissions. If you do, have each member sign in individually rather than signing in once for everyone, so each person works through their own nCino sign-in.

Perplexity uses a connector only when it's turned on for that chat. Turn it on under Sources with the + button in the prompt box, or @mention it.

See Perplexity's Adding Custom Remote Connectors.

Step 3: Connect your nCino environment

The assistant reaches your data through a registered Salesforce sandbox org. Production orgs cannot be registered at all, so they are never reachable.

Which path you take depends on your nCino release.

On nCino 2026_08 or later, with nFORCE 2.3137 or higher, your administrator enables it in Setup. Admin Digital Partner is a sub-feature under nCino AWS Configuration.

On earlier releases, nCino provides a script that completes the registration instead.

This process is run during the onboarding discussion with your CSM.

Network access. Some AI assistants connect to the nCino MCP from the vendor's cloud rather than from your computer. If your network team filters outbound traffic, they may need to allow the addresses your assistant connects from. Check your assistant's documentation, and ask them to confirm before you roll out.

Set up approval for changes

Before you make the connector available to your administrators, decide which actions require someone to approve them first. That setting lives in your AI assistant, not in nCino. See Limitations.

The change tools are off by default. If your institution wants the connector to be able to make changes, ask your Customer Success Manager and nCino turns them on for you.

Confirm it is working

After step 2, check that your assistant is connected and signed in. This check needs no org:

"Using nCino, show me what's in nCino release 2026_08, including its package versions."

After step 3, check that your org is reachable:

"List my registered nCino sandbox orgs, then show me the active validation rules on the Loan object in the one I pick."

If both come back with real content, your assistant is connected and signed in correctly.

A healthy setup lists your registered sandboxes, waits for the administrator to choose one, and then returns the validation rules from that org. Choosing an environment is always an explicit step. The connector never guesses which org to act on.


Did this page help you?