Sprad is the applicant tracking system with an open API and a native MCP connection, the AI-first ATS you run from your own assistant while you confirm each action. Ask Claude, ChatGPT or your own agent to read the pipeline and act on it, and hiring happens inside a system built for exactly that.

For a recruiter, that means you stop being the copy-paste layer between the AI you trust and the tool where hiring happens. Your assistant works against live candidate data and hands each decision back to you to approve. Very few platforms can do this yet, which is what makes the definition worth getting right.

The AI you already use everywhere else has, until now, been locked out of the one system where your hiring happens.

What is an ATS with an open API and MCP?

An ATS with an open API and MCP is a hiring system whose data and actions are exposed so an outside AI assistant can operate it directly, from reading the pipeline to acting on it. The open API publishes the core objects a recruiter works with every day, and MCP is the layer that lets your AI find those objects and use them in plain language.

The API is the substance here. A genuinely open ATS API exposes the standardized objects a recruiter touches daily, from open jobs and candidate records through to interviews and scorecards, with both read and write access. Read access lets an assistant summarize a pipeline, while write access is what lets it advance a candidate or send an interview invite. Take away the write scopes and an agent can only describe your hiring, never move it.

Very few hiring platforms actually offer this today. Most legacy vendors are held back by older data models and slower release cycles, so native support for your own AI stays a real differentiator through 2026. Sprad's free AI ATS is built for it from the ground up, EU-hosted and human-in-the-loop by design, with a core that stays free forever and optional AI modules on top.

What is MCP (Model Context Protocol), and why does it change hiring?

MCP, the Model Context Protocol, is the open standard that lets your AI assistant read your systems and act inside them. Anthropic open-sourced it in November 2024, and that read-and-write ability is what turns an ATS from something you click through into something your assistant runs for you.

Model Context Protocol (MCP): an open standard that connects AI assistants to the tools and data they need, so they can both retrieve information and perform tasks. It is widely described as "a USB-C port for AI applications."

MCP spread fast. Within about a year it became the vendor-neutral way to connect AI agents to tools, with more than 10,000 active public servers and adoption across every major assistant, before it was handed to the Linux Foundation's Agentic AI Foundation in December 2025.

For hiring, the practical result is direct. With your ATS connected through MCP, you can run day-to-day recruiting in plain language inside Claude or ChatGPT, from building a shortlist to drafting the outreach that goes with it, all against live ATS data with no copy-paste or export. Sprad's own AI HR agent, Atlas, is the piece that executes those natural-language workflows for you.

How do you run hiring from Claude or ChatGPT?

You run it as a conversation, and you stay the one who approves each step. A single request like "show me the top five candidates for the backend engineer role and invite them to a voice interview" turns into a sequence your agent proposes and you confirm before anything happens in the system.

Here is how that request plays out with Sprad connected to your assistant:

The voice interview in that example is one of Sprad's built-in modules, a structured AI voice screening that produces a report with quoted evidence back in the ATS. The confirmation step carries real weight too. MCP is built so the client asks you to grant permission before a tool runs, and best practice is to require explicit sign-off for any write action, exactly the sensitive steps like sending an invite or rejecting a candidate.

MCP control vs. a closed ATS integration: what's the difference?

The core difference is who decides what the software can do. A classic integration runs along a fixed path a developer wired in advance, while MCP lets your AI discover an ATS's tools and call them at runtime, in plain language. MCP layers on top of the API rather than replacing it, so the API keeps doing the real work and the agent gets a standard way to reach it.

The distinction turns into a business risk the moment your AI ambitions outpace your ATS. In a closed ecosystem you wait for the vendor to ship each feature and you use whatever AI they bolt on, on their schedule, with no way to bring the assistant your team already relies on. As autonomous agents take on more of the workflow, that lock-in becomes a real liability. More than half of talent leaders plan to add autonomous AI agents to their teams in 2026, and recruiting is already the single most common use of AI in HR, so the systems that stay closed will feel the gap first.

An open ecosystem points the other way. Sprad connects with the systems recruiting teams already run, from HR suites like Personio and Workday to sourcing channels like LinkedIn and Indeed, and it fits alongside the other major ATS platforms teams depend on. We unpack why the integration ecosystem decides everything and what agentic HR software actually delivers in separate guides.

Is an ATS with an open API and MCP secure and EU AI Act compliant?

Yes, when the right controls are designed in from the start. The genuinely new risk MCP adds is prompt injection and over-permissioned agents, where a manipulated instruction pushes an assistant to do more than it should. The defenses for it are well understood, and Sprad applies them by default.

The controls that make it safe: least-privilege scopes so an agent only touches what it needs, explicit human confirmation before any write action, a complete audit trail of every AI step, and EU-hosted data that never leaves the region.

Compliance is the other half, and in Europe it is not optional. The EU AI Act classifies AI used in recruitment as high-risk, which brings duties around traceable activity logging and genuine human oversight. The high-risk obligations for hiring now apply from 2 December 2027 after the AI Omnibus deferral, with transparency duties arriving in August 2026, so teams have a real window to get their setup right.

Two rules make the human role non-negotiable. Article 14 of the AI Act says a person must be able to interpret a hiring system's output, override it, or stop it entirely, and must be protected against automation bias. GDPR Article 22 gives candidates the right not to be subject to a hiring decision based solely on automated processing, and the courts have confirmed that a rubber-stamp review does not count. A confirm-before-you-act model is how you stay on the right side of both.

Sprad is built for exactly this setup. The platform is EU-hosted and compliant with both GDPR and the EU AI Act, and its screening is bias-audited, with the evidence behind every decision. It stays human-in-the-loop by design and never trains models on your data, and its audit trail records every AI action next to the person who approved it, which is exactly the evidence an auditor or works council will ask to see.

When the ATS becomes what your AI operates

Once your AI can operate the ATS, something quietly shifts. The system fades into the background and your assistant works it for you. The interface moves from clicking through forms to describing what you want, and your judgment lands where it matters, approving the decisions. You stop asking which ATS has the slickest screen and start asking which one your AI can safely operate.

Both the open API and the MCP connection matter more than any single feature. They decide whether the AI you already trust can reach your hiring at all, and whether it does so under permissions, confirmation and a clean audit trail. In Europe, that governance is what keeps plain-language hiring both fast and defensible.

Sprad brings the two together in one place: an AI-first ATS with a free-forever core and optional AI modules, plus Atlas ready to run your hiring from Claude, ChatGPT or your own agent. You can connect your own AI to Sprad's ATS and keep every action behind your confirmation.

Frequently asked questions

What is MCP in an ATS?

MCP in an ATS is the connection layer that lets an outside AI assistant operate your hiring system directly. It exposes the ATS's read and write actions to tools like Claude or ChatGPT, so the assistant can search candidates, build shortlists and move people through stages in plain language, with your confirmation on each step.

Can I control an ATS with ChatGPT or Claude?

Yes, if the ATS exposes an open API and an MCP connection. With Sprad, you ask Claude, ChatGPT or your own agent to read your pipeline and propose actions, then you confirm before anything is sent or changed. The AI does the work, and you keep the decisions and the approval.

Which ATS supports MCP?

Sprad is the AI-first ATS built around an open API and a native MCP connection, so you can run hiring from your own AI assistant. Native MCP support is still rare in 2026, mostly limited to AI-native platforms, which makes it a genuine differentiator in the market today.

Is an open-API ATS secure?

Yes, an open-API ATS is secure when it enforces the right controls, above all least-privilege access and explicit human confirmation before any write action, backed by a full audit trail and in-region hosting. The main new risk with AI control is prompt injection, which those safeguards contain. Sprad builds them in by default.

What is the difference between an integration and MCP control?

An integration is a fixed connection a developer builds in advance between two systems along a set path. MCP control lets an AI agent discover a system's tools and call them at runtime in plain language. MCP works on top of the API, so most setups end up using both together.