Turn MCP connects Claude, ChatGPT, Gemini, and custom agents to Turn's screening platform.
Order a check, collect consent, read results, and handle adverse action in plain language. People make the decisions. Turn's controls run on every step.
Customer Access is live for authenticated Turn accounts. Verified Access for new organizations is opening in phases. How access works.
Turn MCP is a hosted Model Context Protocol server that exposes Turn's screening order lifecycle as tools. Any MCP-compatible agent can list packages, create an order, request consent, check status, retrieve results, and start adverse action, acting on behalf of an authenticated Turn account.
Hiring work is moving into recruiting copilots, ATS assistants, and internal agents. Those agents need to act, not just look things up: trigger a check, read the result, and route the next step inside one conversation.
Most background check APIs were written for engineers writing code, not for agents working in natural language. Turn MCP closes that gap. The order, consent, and resolution logic that runs Turn's core platform is published as a set of typed tools with Turn's eligibility checks, confirmations, and data masking built in.
Start a standard background check for Maria Chen and send her the consent request.
The order is open and waiting on Maria's authorization. I'll report status when results post. Median turnaround is about six hours.
Turn MCP exposes the full order lifecycle as callable tools, not a read-only reporting layer. Everything listed here is live in production for Turn customers.
Many screening MCP servers only let an agent read data that already exists. Turn MCP includes write tools that create orders, send consent requests, and start FCRA adverse action workflows, so an agent can carry a screening task from start to finish instead of reporting on one already in motion.
A single Streamable HTTP endpoint serves Claude Desktop, ChatGPT, Cursor, and a LangChain agent alike. There is no per-platform SDK and no duplicated integration work.
Agent-initiated orders run on the same infrastructure as every other Turn order, with a six-hour median turnaround. Going through MCP adds no latency.
Turn MCP is a standards-based server, not a plugin for one marketplace. It speaks the protocol every major agent platform has adopted, so the connection you set up once keeps working as the ecosystem changes.
Claude Desktop, web connectors, and Claude Code.
ChatGPT MCP connectors and the OpenAI Agents SDK.
Gemini's MCP-aware agent tooling and the Agent Development Kit.
Run screening actions from a developer's coding agent.
Add Turn tools to any custom LangChain agent graph.
Place screening steps inside low-code automation workflows.
Call Turn tools from Bedrock-hosted agent runtimes.
Any framework that speaks MCP over Streamable HTTP works without changes.
An API asks your developers to build and maintain client code for each application. A read-only MCP server lets an agent summarize what already happened. Turn MCP lets an agent complete the work, with Turn's controls in the loop.
| Capability | Turn MCP | Read-only screening MCP server | Traditional background check API |
|---|---|---|---|
| How the agent interacts | Plain-language tool calls | Plain-language tool calls | Custom code against REST endpoints |
| Read report status and results | Yes | Yes | Yes, with code |
| Create a new screening order | Yes | No | Yes, with code |
| Send a candidate consent request | Yes | No | Yes, with code |
| Start an FCRA adverse action workflow | Yes | No | Depends on the vendor |
| Human confirmation on sensitive actions | Required, enforced by Turn | Not applicable | Left to the integrator |
| Integration effort | Add endpoint, authenticate with OAuth | Add endpoint, authenticate | Build, test, and maintain a client per app |
| Works across agent platforms | One endpoint, any MCP client | One endpoint, any MCP client | Separate build per application |
Comparison reflects publicly documented screening MCP servers and background check APIs as of September 2026. Read-only servers generally expose tools to look up candidates and retrieve existing reports. Vendor capabilities change; verify against current documentation.
Turn MCP places real screening orders and touches consumer data, so access is deliberately controlled. Existing Turn customers can connect today. New organizations join through a verification process that is being completed now.
For existing Turn customers with authenticated accounts
Your organization has already completed Turn's onboarding, credentialing, and contracting. Turn MCP extends that same account to your agents, with the permissions and packages you already have. There is no separate agreement and no new login.
For organizations new to Turn
New organizations will be able to connect an agent and begin screening after passing Turn's two assurance layers, described below. Verified Access opens in phases, beginning with organizations on the access list, and brings the same controls that apply to Turn's existing customers into the agent flow.
A background check is a regulated consumer report. Before an organization that Turn has not met can place one through an agent, Turn confirms who they are and that they are in good standing to pay for it. Both layers run before any screening tool becomes available, and both are in progress now.
Verifies that the organization is a real, eligible business and that the person connecting an agent is authorized to act for it. This is how Turn establishes permissible purpose before the first order.
Confirms a valid payment method and account standing before an agent can create an order, so every order is tied to a verified, accountable customer.
Letting software create legal screening actions raises the bar. These controls are enforced by Turn on every Turn MCP call today, regardless of which agent is on the other end.
Turn is a Consumer Reporting Agency subject to the FCRA. Turn MCP supports FCRA-required workflows but does not replace the employer's own obligations, including establishing permissible purpose, obtaining authorization, and issuing adverse action notices. Consult qualified legal counsel for requirements specific to your jurisdiction and role.
Every request carries an OAuth token scoped to one Turn account. No static API keys are shared with the model.
An order, consent request, or adverse action only proceeds if the account and candidate meet Turn's eligibility rules.
Consent, adverse action, and any decision step require an explicit typed confirmation from an authorized person. The agent cannot supply it on its own.
Every call is recorded against the account, so an administrator can see what an agent did and when.
Turn MCP does not adjudicate. Approve, reject, and withdraw decisions stay with an authorized human user.
The protocol handshake takes minutes. Most of the time goes to reviewing the guardrails with your team before you point a production agent at real applicant records.
Paste the Streamable HTTP URL into Claude Desktop, ChatGPT, Cursor, or your framework's MCP configuration. The agent discovers the tools and their schemas automatically. No manual wiring.
Turn MCP uses OAuth with dynamic client registration, as the MCP specification describes. The agent acts on behalf of your Turn account, and the model never holds a static key.
Walk through order creation, consent, status, and results against sandbox data. Confirm the typed-confirmation steps behave the way your compliance team expects.
Switch the endpoint to production. The same eligibility checks, confirmations, and masking apply, and future control layers arrive without asking you to reconnect.
What is live today is a complete, production integration for Turn customers. The work in progress widens who can use it and deepens the controls around it.
Verified Access opens to new organizations in phases, with verification in front of every one of them. The access list is open now.
Broader access arrives with stronger controls, not looser ones. Each layer is designed to be auditable and to sit in front of the screening tools, not beside them.
Coverage of Turn's workflows through MCP will keep growing. Connections built today carry forward without a rebuild.
Direct answers for buyers, developers, and compliance reviewers. For implementation detail, see the MCP Knowledge Library.
Turn MCP is a Model Context Protocol (MCP) server operated by Turn, a Consumer Reporting Agency subject to the FCRA. It exposes Turn's background screening workflows, including creating orders, requesting candidate consent, retrieving results, and starting adverse action, as tools that any MCP-compatible AI agent can call.
Agents connect over a single hosted endpoint and act on behalf of an authenticated Turn account, with Turn's eligibility checks, confirmations, and data masking applied to every call.
The Model Context Protocol is an open standard for connecting AI agents to external tools and data. An MCP server publishes a set of tools with typed inputs, and any MCP client, such as Claude, ChatGPT, Gemini, or Cursor, can discover and call those tools during a conversation.
Turn MCP uses this standard so one integration works across agent platforms instead of requiring a separate build for each.
Turn MCP works with any client that implements the Model Context Protocol over Streamable HTTP. That includes Claude (Desktop, web connectors, and Claude Code), ChatGPT and the OpenAI Agents SDK, Gemini agent tooling, Cursor and other IDE agents, LangChain and LangGraph, n8n, Amazon Bedrock agents, and custom agents you build yourself.
Turn does not favor one platform. The same endpoint serves all of them.
Turn is a Consumer Reporting Agency subject to the FCRA, and Turn MCP runs the same order, consent, and adverse action workflows as Turn's platform. Every sensitive action requires an authenticated account, an eligibility check, and an explicit typed confirmation, and candidate identifiers are masked in tool responses.
The agent calls tools; it does not make hiring decisions. Employers remain responsible for their own FCRA obligations, including permissible purpose and adverse action notices, and should consult qualified counsel.
No. Turn MCP is a hosted server, so there is no SDK to install and no client code to write. You add Turn's MCP endpoint to your agent's configuration, authenticate with OAuth, and the agent discovers the available tools automatically.
Existing Turn API and ATS integrations continue to work alongside it.
Turn MCP is available now to existing Turn customers with authenticated accounts. This is called Customer Access. Organizations that are new to Turn can request Verified Access, which opens in phases once Turn's Identity Assurance and Billing Assurance layers are live.
Verified Access for new organizations is in progress and will open in phases, starting with organizations on the access list. Turn is completing two verification layers first: Identity Assurance, which confirms the organization and the people acting for it, and Billing Assurance, which confirms payment standing before any order is placed.
Request access to be notified when your organization can connect.
A background check API requires developers to write and maintain code for each application that orders or reads checks. Turn MCP publishes the same screening actions as tools an AI agent can call in plain language, with authentication, eligibility checks, and confirmation steps handled by Turn.
It also differs from read-only screening MCP servers, which let an agent summarize existing reports but not create orders, send consent requests, or start adverse action.
An agent can list screening packages and add-ons, create a screening order, send and check candidate consent, track order status, retrieve results, and start an FCRA adverse action workflow.
An agent cannot approve, reject, or withdraw a candidate on its own, cannot bypass consent, and cannot see full Social Security numbers or dates of birth. Sensitive steps require a typed confirmation from an authorized person.
Turn MCP authenticates every request with OAuth and scoped tokens, isolates data to the authenticated account, and masks personal identifiers such as full SSNs and dates of birth before a response leaves Turn.
Tool calls are logged to the account, and no static API keys are shared with the model.
Put a background check inside the conversation your team is already having. Turn customers can connect an agent today with sandbox and production endpoints and full MCP documentation. New organizations can join the access list and be first through Verified Access.
Download SOC 3 Report
Get a high-level overview of Turn’s security controls and compliance practices.
