Workable Agent and Workable MCP Server: A complete support guide

Hiring teams are under pressure to move faster with fewer resources, and AI is quickly becoming the difference maker in how recruiters source, screen, and engage candidates. During the Workable Agent and MCP 101 webinar, our product and sales teams walked through exactly how the Workable Agent and the Workable MCP Server work, and three customers, H2 Health, Belong Villages, and CacheFly, shared how each tool has changed their day-to-day recruiting.

This guide brings that session together in one place. Whether you are evaluating the Workable Agent for the first time, rolling out your MCP integration, or looking for proof points to bring back to your team, use this as your reference.

What is the Workable Agent?

The Workable Agent is not a single tool. It is a series of autonomous agents that work together, all built around one core concept: the ideal candidate profile (ICP). The ICP acts as the brain behind every open role, and it is what allows Workable’s agents to evaluate, source, and engage candidates consistently throughout the hiring process.

There are four agents that build off the ICP:

• Job Brief Agent. This agent helps define the role from the start, going well beyond a standard job description. Through a natural-language, back-and-forth conversation (or by reading an existing job description), it produces up to 14 different criteria, categorized as must-haves, nice-to-haves, and disqualifiers. Recruiters can edit, add, or remove criteria before the ICP is finalized, and can control what appears publicly in a job post versus what is used only internally.
• Sourcing Agent. Workable’s database of 400 million candidate profiles is searched and continuously updated against the ICP. In practice, this has produced a 10% open rate and a 4% passive completion rate, roughly double what is typical in the industry.
• Screening Agent. Every inbound applicant is scored against the ICP as soon as they apply, rather than being reviewed in batches once a week. Each score comes with a documented, written explanation of why a candidate matched or did not.
• Engagement Agent. High-touch candidates are engaged automatically. Outreach is identified and, when a candidate responds, they are moved into a screening conversation with a full, searchable, and downloadable transcript.

How AI Credits Work

Every Workable ATS account includes 3,000 complimentary AI credits, enough to try the agent across roughly three roles before deciding on future investment. Credits are consumed based on the level of effort involved:

• Evaluation (scoring a candidate against the ICP): 1 credit
• Sourcing (outreach to a passive candidate): 2 credits
• Chat (a full engagement conversation): 10 credits

An AI credit report is available inside the platform so teams can track balances, see how credits are allocated across chat, evaluation, sourcing, and re-evaluation, and forecast usage over time.

Trust, Control, and Compliance

Because AI in hiring naturally raises questions about fairness and compliance, the Workable Agent is built around five principles:

1. Transparency. Every candidate score comes with a documented reason. Candidates are always informed when they are interacting with AI.
2. Control. Recruiters can turn individual agents on or off per role and override any AI decision.
3. Data integrity. Evaluations are based solely on the ICP, not on a company’s historical candidate data, which avoids inheriting past bias.
4. Bias mitigation. Every candidate is evaluated the same way, against professional qualifications only, with regular internal audits for model drift.
5. Security and regulation. Workable maintains ISO and SOC 2 compliance, aligns with the EU AI Act, and contributes to AI research and policy work in the EU, alongside GDPR and U.S. state-level compliance.

What is the Workable MCP Server?

The Workable MCP (Model Context Protocol) Server lets AI assistants, such as Claude, ChatGPT, and Microsoft Copilot, connect directly to your live Workable data using natural language, without ever opening the Workable interface. It is available to any customer on a paid Workable plan, and it mirrors the exact permissions a user already has in Workable: if a user cannot see something in Workable, their AI assistant cannot see it either.

Setup takes minutes. From an assistant like Claude, a user searches for the Workable connector, authenticates, and is ready to go. Copilot requires a bit of extra configuration, but none of it requires a developer or IT admin.

The MCP Server launched in May with 38 tools. By July, that had grown to 94 tools, covering both read and write access across:

• ATS data: jobs, candidates, pipelines, and offers
• HRIS data: performance reviews, time off, and calendar

In the webinar’s live demo, a single natural-language prompt pulled a real-time, pipeline-stage breakdown of candidates for an open role, no dashboard refresh required. A follow-up prompt filtered those candidates by how long they had been sitting in the applied stage, the kind of query that would normally require a saved report or a pivot table. From there, three systems were chained together in a single instruction: Workable to pull a list of delayed candidates, Slack to draft and send a message about the pipeline delay, and a calendar tool to schedule a follow-up meeting, all without switching tabs or writing a line of code.

The MCP Server can also power live, interactive dashboards. One example built during the demo, a “talent acquisition command center,” combined open requisitions, pipeline health, candidate sourcing, and hiring velocity into a single view, with the ability to approve candidates or add profile comments directly from the dashboard, writing those actions straight back into Workable.

Ad tech company LoopMe uses the Workable MCP Server to generate a weekly executive report in minutes, a task that previously took hours.

Customer Stories: How Teams Are Using the Workable Agent

H2 Health: 10 Hours Saved Per Recruiter, Per Week

H2 Health is a network of more than 300 outpatient rehabilitation and therapy clinics across the United States, spanning outpatient, senior services, and pediatric divisions. Its recruiting team is lean: three recruiters, each responsible for roughly 100 open positions, supported by additional staff who help with sourcing and administrative tasks.

Carol Ribeiro, H2 Health’s director of recruiting, described the team’s biggest challenge as finding qualified candidates and getting responses in a market where AI-driven outreach has made candidates more desensitized to emails and texts. To counter that, her team leaned back into cold calling, something the Workable Agent’s automated sourcing freed up time to do.

Before the Workable Agent, Ribeiro’s team was, in her words, “flying a little bit blind,” often working from just a name or contact information with no way to gauge who was truly qualified. With Workable’s stack ranking and recommendations, the team now focuses its attention on qualified candidates first, changing the overall hiring strategy and helping the team hire faster.

It took the team less than a day to get comfortable with the Workable Agent, using Workable’s own training videos, and they began using the platform almost immediately. The team has made several hires directly through Workable Agent recommendations and resurfaced candidates.

Asked to quantify the impact, Ribeiro estimated the Workable Agent saves her team at least 10 hours per week per recruiter across roughly 300 open positions.

“AI, it just augments what we’re currently doing, and it just helps us work more efficient and smarter,” Ribeiro said. “It’s not a replacement for somebody, it’s truly just making things more efficient.”

Belong Villages: A Critical Role Filled in 7 Days

Belong Villages is a nonprofit dementia care organization based in the UK with roughly 1,500 employees. Competing for talent in a tight healthcare labor market as a nonprofit, the organization wanted to widen its candidate pool beyond traditional care-sector backgrounds, looking instead for people with the right skills and mindset regardless of prior industry experience.

Using the Workable Agent, Belong Villages filled a critical head of operations role in just seven days, sourcing qualified candidates from outside the care sector who were still strong fits for the role. The process also kept the organization competitive with larger employers: even C-suite and senior-level candidates gave positive feedback on their experience throughout the hiring process.

CacheFly: 930+ Applications Processed in 36 Hours

CacheFly is a fully remote content delivery network (CDN) startup with about 30 employees globally. When its head of HR left the organization, hiring responsibilities landed unexpectedly on the CFO’s desk, on top of an already full workload.

The team needed a way to manage recruiting quickly and without dedicated HR support. Using the Workable Agent, CacheFly processed more than 930 applications for a single role in just 36 hours. Custom knockout questions allowed the agent to automatically eliminate 20% of unqualified applicants, and the result was three roles filled quickly, freeing the CFO to get back to running finance rather than triaging resumes.

Getting Started with the Workable Agent and MCP Server

A few practical starting points, drawn directly from the webinar:

• Use your free credits first. Every Workable ATS account comes with 3,000 complimentary AI credits, enough to test the agent across about three roles before making a larger investment decision.
• Start with training video content. H2 Health’s team got up and running in a few hours by watching Workable’s existing product videos before diving in. If your team learns best by watching, review Workable’s video walkthroughs (including recorded demos like the one referenced in this guide) before your first live role.
• Set your guardrails early. Decide up front which agents you want active on a given role (sourcing, screening, engagement) and how much autonomy to give them, such as auto-advancing candidates who meet 100% of your must-haves.
• Connect your MCP Server. If your team already works inside Claude, ChatGPT, or Copilot, connecting the Workable MCP Server takes minutes and requires no developer support. It is included in every Workable plan.
• Book a workshop. Workable offers one-on-one agent workshops with your account team to help calibrate credit usage and apply the agent to real, live roles rather than a generic demo.

The Bottom Line

Across very different organizations, a 300-plus-location healthcare network, a UK nonprofit care provider, and a 30-person remote startup, the Workable Agent and Workable MCP Server are solving the same underlying problem: recruiting teams need to do more with fewer people, without sacrificing candidate quality or compliance. Whether that means reclaiming 10 hours a week per recruiter, filling a critical role in a week, or processing hundreds of applications in under two days, the pattern across these stories is consistent: automation handles the volume, and recruiters get their time back for the work that actually requires a human.

To see the Workable Agent and MCP Server in action for your team, visit Workable or reach out to your account manager to schedule an agent workshop.