Using Persona with AI agents (MCP Server and OAuth connector for Claude and ChatGPT)

Overview

The Persona MCP (Model Context Protocol) Server lets you connect Persona directly to AI agents and tools, such as Claude, Cursor, or any MCP-compatible application. Powered by the same Persona API your team already uses, the MCP Server exposes Persona’s endpoints in a format that AI systems natively understand.

Persona has two MCP servers. You can connect an AI agent in one of these two ways:

• OAuth connector for Claude and ChatGPT. You sign in with your own Persona Dashboard account using OAuth. You do not need an API key. The assistant works under your own Persona role. See OAuth connector for Claude and ChatGPT.
• API key MCP Server. You connect Claude, Cursor, or any MCP-compatible agent or developer tool with a Persona API key. It gives access to the full Persona API. See API key MCP Server.

Disclaimer: Persona provides tools to help companies address identity challenges, but it is up to each business to determine its risk tolerance and goals, while using Persona appropriately to build its identity programs. When granting AI agents access to your Persona data, ensure your implementation aligns with your organization’s internal security and compliance requirements.

Which option should I use?

Use the MCP server that fits your task.

OAuth connector for Claude and ChatGPT

The Persona connector lets you ask questions about your Persona data from inside Claude (claude.ai) or ChatGPT. It is a remote MCP (Model Context Protocol) server at https://withpersona.com/mcp. You sign in with your own Persona Dashboard account using OAuth and approve access on a Persona consent screen. You do not need an API key.

The assistant can search, fetch, and list records. It can also make a small set of changes in Persona. It cannot delete anything.

What can the assistant do?

Once connected, the assistant can read the following kinds of records, among others:

• Inquiries and Inquiry Templates
• Accounts
• Cases
• Verifications
• Reports
• Transactions
• Workflows
• Lists

It can also list your Persona environments.

What can the assistant change?

The assistant can do these actions:

• Create an Inquiry from an Inquiry Template. You can ask for a one-time link. Anyone who has the link can complete the Inquiry, so share it only with the intended person.
• Add a comment to a Case.
• Add a tag to a Case, or remove a tag from a Case.
• Set the status of a Case.
• Assign a Case.
• Approve or decline an Inquiry.

Each change runs as you, under your Persona role. Persona records each change in your organization’s audit log. Claude and ChatGPT ask you to confirm before they take an action.

Connecting Claude or ChatGPT

When you ask for a change for the first time, the assistant can ask you to reconnect and approve write access on the Persona consent screen.

If you cannot connect, contact your Persona account team or Persona support.

Permissions and access

• The connector uses your own Persona role. The assistant can read only what you can read in the Dashboard.
• The assistant can make a change only if your role lets you make it in the Dashboard.
• You approve write access on the Persona consent screen.
• Results are limited to your organization.
• Some resources, such as Inquiries, belong to an environment. Ask the assistant to list your environments first, then say whether you mean production or sandbox.

Example prompts

• “List my Persona environments.”
• “Show my 5 most recent inquiries in sandbox.”
• “Summarize the verifications on inquiry inq_….”
• “Search for the account for jane@example.com.”
• “Add the comment ‘Documents checked’ to case case_… and set its status to Resolved.”
• “Create an inquiry from template itmpl_… in sandbox and give me a one-time link.”

Disconnecting

You can disconnect Persona at any time in Claude or ChatGPT’s connector settings. After you disconnect, the assistant stops using your Persona access.

Data handling

Data that the connector returns to the assistant is sent to the AI provider, and the provider’s terms govern how it handles that data. Review those terms before you connect, and read the Persona privacy policy to learn how Persona handles data.

API key MCP Server

The API key MCP Server is at https://mcp.withpersona.com. It authenticates with the API keys you already have in your Persona Dashboard.

What capabilities does this unlock?

The MCP Server bridges the gap between your identity data and your AI ecosystems. By exposing over 190 standard API actions as native AI tools, it allows businesses to build highly customized, agentic workflows without writing extensive custom integration code. Browse all resource groups in the Persona API Reference.

Key capabilities unlocked include:

• Automating data exploration: Enable AI agents to instantly cross-reference fraud signals, pull Reports, or query data to identify discrepancies between Persona and your internal databases.
• Streamlining manual review: Give internal AI assistants secure, read-only access to Cases and Accounts. This allows operations teams to use AI to summarize Verification data, detect anomalies, and make faster, more informed decisions.
  • Note: Persona also offers native Case Review Agents built directly into Cases to assist with manual review out-of-the-box. The MCP Server is ideal if you want to integrate Persona data into your own external AI ecosystems instead.
• Accelerating development: Allow engineering and product teams to use AI coding assistants to seamlessly prototype configurations, create Workflows, and test endpoints.
• Full platform access: The server supports all major Persona API resource groups, including Inquiries, Verifications (such as Government ID, Selfie, and Database), Reports, Transactions, Webhooks, and user management.

Authentication and security

API Keys

Authentication uses the exact same API keys you already have configured in your Persona Dashboard. Pass your key in your request header using one of the following formats (the server normalizes both to Bearer):

• Authorization: Bearer <your-key>

The MCP Server strictly respects all existing API key scopes and permissions. Your AI agent can only perform actions explicitly permitted by the key provided. Keep in mind that the MCP Server exposes all available API actions as tools to your AI agent, regardless of your key’s permissions. If the agent attempts an action outside the key’s scope, Persona’s API will return a permission error. For this reason, we recommend creating a dedicated key for each agent with only the permissions it needs. To configure or review your key’s permission scopes, see API Keys in the Persona documentation.

Stateless design

The server is stateless and requires no session ID. Each request is fully independent, meaning no warm-up or handshake is needed, it is safe to load-balance, and no cleanup is required when a session ends. It passes requests directly to Persona’s API and returns the response. No session data, conversation history, or credentials are stored by the MCP Server between requests.

Example: if an AI agent fetches an Inquiry in one request and then retrieves a related Verification in a follow-up request, the server treats these as two fully independent calls. There is no shared context between them on the server side. Your agent or orchestration layer is responsible for passing any relevant data (such as IDs or tokens) between steps.

Connecting to the MCP Server

To connect, configure your AI tool or agent to use the following production endpoint and provide your API key: https://mcp.withpersona.com

Prompting best practices

AI agents perform best with clear, specific instructions. We recommend the following best practices when building prompts:

• Be explicit about resource identifiers: Always include the exact token type and value. For example, instruct the agent to “create an Inquiry using template itmpl_xxx” rather than simply “create an inquiry.”
• Scope your API key: Use a read-only key if the agent only needs to pull and review data. This prevents the agent from accidentally creating or modifying records.
• Paginate intentionally: Most list endpoints return cursor-paginated results. Explicitly ask the model to “fetch the next page using the cursor from the previous result” rather than letting it loop unguided.
• Pin your API version: Use the Persona-Version optional header (e.g., Persona-Version: 2025-12-08) to ensure consistent behavior as the API evolves. If omitted, it defaults to the server default.

Frequently Asked Questions (FAQ)

How do I connect the MCP Server to ChatGPT, Claude, or Grok?

For Claude or ChatGPT, you can use the OAuth connector at https://withpersona.com/mcp. You sign in with your Persona account and you do not need an API key. See OAuth connector for Claude and ChatGPT.

The API key MCP Server authenticates with an API key. It is not listed in any of these products’ official connector directories. In each one, you add https://mcp.withpersona.com as a custom (bring your own) MCP connector and authenticate with an existing Persona Dashboard API key:

• Claude (Anthropic): In Claude’s connector settings, add a custom connector. Custom connectors are primarily OAuth-based, but a beta “static header” option lets an org admin configure a fixed header (Authorization or x-api-key) with your API key as the value, shared across the org. Alternatively, via the Anthropic Messages API, the mcp_servers config takes an authorization_token field sent as a Bearer token, which accepts a plain API key.
• ChatGPT (OpenAI): With Developer Mode enabled (on Business or Enterprise workspaces, an admin may need to enable creating custom MCP connectors first), add a custom connector in ChatGPT’s connector settings and choose API key as the authentication type. Via the Responses API, the authorization field on the mcp tool is sent as Authorization: Bearer <value>.
• Grok (xAI): In Grok’s connector settings, add a custom connector, enter the MCP server URL, and complete the API key authentication. Via xAI’s API, the remote MCP tool config takes an authorization field (sent as a Bearer token) or a generic headers field for other header names.

What are some common use cases for the MCP Server?

Because the MCP Server exposes the full capabilities of the Persona API, it can be used for a wide variety of agentic workflows. Common customer use cases include:

• Assisting Operations Teams: Building internal AI tools that pull read-only Verification data to assist agents during the manual review process in Cases.
• Data Cross-Referencing: Using an AI agent to identify discrepancies between values in Persona (like fraud signals or passed Inquiries) and an internal database.
• Report Generation: Using AI agents to securely pull and format data from Persona Reports or aggregate bulk data for visualization.
• Integration and Testing: Empowering engineering and product teams to easily test API endpoints, update Template configurations, or prototype Workflows.

How are rate limits and errors handled?

The MCP Server uses the exact same rate limits as the standard Persona external API. If your agent hits a rate limit, it will receive a 429 error, which should be handled with exponential backoff. All error responses follow the standard Persona error envelope.