TalentGuard Brings WorkforceGPT Workforce Intelligence to Claude and Other AI Agents Through MCP
New connector lets HR teams and managers score job descriptions and build structured role profiles, with skills, proficiency levels and target expectations, inside Claude and other AI agents using a free WorkforceGPT account
AUSTIN, Texas — September 23, 2026 — TalentGuard today announced that WorkforceGPT, its workforce intelligence engine, is listed in Claude’s Connectors Directory and available to other AI agents through the Model Context Protocol (MCP). HR teams and managers can now score job descriptions and build structured role profiles, complete with skills, proficiency levels and target expectations, without leaving the AI tools they already use.
WorkforceGPT works in Claude on the web, Claude Desktop and Claude Code, where users add it from the directory with nothing to configure. Any other MCP-compatible client can connect with the server URL alone. A WorkforceGPT account is free, and organizations do not need to be TalentGuard customers to use it.
Ask three general-purpose AI systems what skills a role requires and you may get three different answers. That may be acceptable for brainstorming, but it is not a reliable foundation for decisions about people. Those decisions depend on a consistent definition of what the role does, which capabilities matter, what proficiency looks like and what level the role requires. AI can write a job description. WorkforceGPT builds the workforce standard behind the job, so an agent can work from a governed definition of the work instead of inventing one.
“AI is making it dramatically easier to create content, but workforce decisions require more than content. They require trusted data about the work itself,” said Linda Ginac, co-founder and CEO of TalentGuard. “HR and business leaders need to know what capabilities a role requires, what good looks like at each proficiency level and how that connects to development, mobility and succession. We aren’t building another AI assistant. We would rather be the intelligence behind every agent in the enterprise than one more assistant competing for attention inside it. Being in Claude’s Connectors Directory puts that intelligence at the point of work, and anyone can try it today with a free account.”
Two Functions Available Today
Through the connector, an agent can currently perform two functions:
Score a job description. A user pastes in an existing job description and asks the agent to assess it. WorkforceGPT returns a score, explains what the score means and identifies the information the source document is missing. The assessment measures how much of a structured role profile the document can actually support and names what to strengthen rather than filling the gaps with inference. The result comes back in the conversation.
Build a structured role profile. Given a job title and organization, WorkforceGPT builds the role profile behind it: a description, responsibilities, required skills, qualifications, proficiency expectations and aligned learning from LinkedIn Learning or Skillsoft. Each skill carries its importance, required proficiency level and behavioral indicators. The output is structured data an agent can put in a table, not prose that has to be reformatted. Skills taxonomy construction and continuous market calibration remain capabilities of the WorkforceGPT platform rather than functions available through the connector.
Knowing that a role requires a skill does not tell an organization how capable someone needs to be to perform the role well. By defining proficiency levels and setting a target level for each skill, WorkforceGPT gives organizations what they need to assess readiness, identify capability gaps, guide development and compare talent consistently across roles.
Skills in a WorkforceGPT profile are grounded in labor-market data, including sources such as the O*NET database, or, when suitable occupational data is unavailable, in the organization’s own job descriptions. TalentGuard’s workforce-specific models then generate and structure the rest of the profile around that foundation. This addresses one of the biggest risks of using general-purpose AI for talent management: plausible-sounding workforce data that is not grounded in an organization’s actual work or an established labor-market source.
Generation Stays Separate From Approval
The connector produces draft role profiles. It cannot publish, approve or write to an organization’s job architecture. Publishing requires a named approver in TalentGuard’s Intelligent Role Studio, where each approved version is recorded, so organizations can bring WorkforceGPT into everyday AI tools without allowing an agent to change the workforce standard on its own.
Access is controlled at the WorkforceGPT account level. An authorized user approves the connection in the browser, under scopes shown on the consent screen, and every request after that runs within that account’s permissions and usage allowance. There is no master key. The connector exposes defined WorkforceGPT functions rather than open access to the TalentGuard platform, and its current capabilities work with roles and job descriptions, not individual employee records.
Building the First Five Links
Every major talent process depends on the same chain of workforce intelligence:
Role → Responsibilities → Skills → Proficiency Definition → Target Level → Employee Capability → Gap → Development → Readiness
WorkforceGPT produces the first five links. Mobility, development, succession, assessment and workforce planning all depend on them. Yet most organizations hold those links as incomplete job descriptions, inconsistent skill names and no common proficiency framework, spread across HR systems, spreadsheets and documents.
“Talent systems are only as useful as the workforce data underneath them. Without it, even the best of them (ours included) are empty vessels of functionality,” said Frank Ginac, co-founder and Chief Technology & AI Officer at TalentGuard. “A skills list alone doesn’t fill them. Organizations need to know which capabilities matter, how proficiency is defined and what level each role requires. WorkforceGPT builds that foundation from labor-market data and the organization’s own job descriptions, and MCP puts it inside the AI agents people already use. The agent can draft a role profile, but it can’t publish one. That step belongs to a person, because we don’t take the ‘human’ out of human capital decisions.”
For HR, that means one standard for jobs, skills, proficiency levels and target expectations, used across talent applications, AI experiences and workforce processes. For managers, it means useful answers without becoming a job architect: what capabilities a role requires, what someone at the next proficiency level should demonstrate or where a job description falls short.
Built to Outlast the AI Tool of the Moment
Because MCP is an open protocol, organizations can connect multiple AI applications to the same WorkforceGPT service instead of maintaining separate role definitions in each one. Unlike a traditional API integration, MCP lets an approved agent find the right WorkforceGPT capability in the moment and use it to answer the question in front of it, with no developer between the person asking and the answer. The interface can change while the workforce standard stays the same. The role and skills data WorkforceGPT produces can support TalentGuard’s talent management applications and, depending on each organization’s implementation and integration approach, other enterprise systems such as SAP SuccessFactors, Workday and PeopleSoft.
Availability
WorkforceGPT is available now in Claude’s Connectors Directory. Other MCP-compatible clients can connect using the server URL, https://workforcegpt.ai/mcp, which also hosts setup guides, a tool reference, worked examples and a description of scopes and limits. A WorkforceGPT account is free with a verified email address and includes role generations, so teams can score a real job description and generate a real role profile before engaging with TalentGuard. A sample role profile is available here. TalentGuard customers use the same engine at enterprise scale, with governance and approval running through the Intelligent Role Studio.
About WorkforceGPT
WorkforceGPT is TalentGuard’s proprietary workforce intelligence engine for building and maintaining structured role and skills data. Unlike a general-purpose language model prompted to generate HR content, WorkforceGPT uses a mixture-of-experts architecture of workforce-specific models trained and calibrated around labor-market information, role taxonomies, industry competency frameworks and organizational skill standards. It generates job descriptions, responsibilities, skills, qualifications, proficiency frameworks, target proficiency expectations and learning alignment. Outputs are designed for review, versioning and subject-matter-expert approval before they are used in talent management processes, and for TalentGuard customers that approval runs through the Intelligent Role Studio. WorkforceGPT is the foundation of TalentGuard’s talent management platform, and the data it produces can also be used with other enterprise HR systems. Create a free account.
About TalentGuard
TalentGuard is a workforce intelligence company that helps organizations understand the work they need, the capabilities their people have and the gaps they need to close. Its platform connects role architecture, skills intelligence, talent assessment, development, career mobility and succession planning so HR and business leaders can make workforce decisions using consistent, governed data. TalentGuard’s approach, Enterprise Skills Trust and Readiness Intelligence (ESTRI), treats skills data like any other governed enterprise asset: an approved standard, verified evidence and a decision trail that can be produced on demand.
TalentGuard is headquartered in Austin, Texas. Learn more at www.talentguard.com.
Claude is a trademark of Anthropic, PBC. O*NET® is a trademark of the U.S. Department of Labor, Employment and Training Administration. All other trademarks are the property of their respective owners.
Media contact: Leo Murphy, Marketing at info@talentguard.com
