OJCP v0.1: An Open Protocol for Agent-Consumable Job Data

OJCP enables AI agents to automate job discovery and application

The Open Job Context Protocol (OJCP) v0.1 is a standardized framework designed to allow AI agents and job providers to communicate using a structured, discoverable, and privacy-respecting language. By building on the Model Context Protocol (MCP) and maintaining interoperability with schema.org, OJCP provides a consistent way for agents to search for jobs, assess candidate fit, and initiate applications without relying on unstructured web scraping.

Built on the Model Context Protocol (MCP)

OJCP is native to the Model Context Protocol (MCP), meaning its tools are valid MCP tools. Any MCP-compatible client can utilize OJCP's standard toolset to perform job-related actions.

Standard MCP Tools

OJCP defines six standard MCP tools. While providers must implement search_jobs to be compliant, the other tools are recommended for full functionality:

  • search_jobs: The primary tool for finding opportunities.
  • get_job_detail: Retrieves detailed information about a specific role.
  • begin_application: Initiates the application process.
  • check_application_status: Allows agents to monitor the progress of a candidate through the hiring pipeline.

Discovery and Manifests

To enable automatic discovery, providers host a manifest at /.well-known/ojcp.json. This manifest allows agents to automatically identify the provider's capabilities, available tools, and supported application paths.

Core Schemas and Data Structures

OJCP defines seven core schemas to ensure data consistency across different job providers:

  • JobPosting: Extends schema.org/JobPosting with agent-specific fields such as ojcp_id, skills_required, experienceLevel, apply_paths, and urgency.
  • CandidateContext: A consent-scoped profile used for personalized results, containing skills, experience years, and employment type preferences.
  • AgentDeclaration: Used for audit trails, requiring the agent to self-identify via agent_id, acting_on_behalf_of, and user_consent_token.
  • VerificationStep: Defines identity verification actions that may require human completion.
  • VerificationProof: A JWS proof containing required claims (iss, aud, sub, iat, exp, nonce) without transmitting PII.
  • VerifierManifest: A discovery document hosted at /.well-known/ojcp-verifier.json specifying verification types and public keys (JWKS).

Normalized Application Taxonomy

OJCP standardizes the fragmented landscape of application mechanisms into a taxonomy that AI agents can reason over. This prevents agents from getting stuck on opaque external redirects or legacy email systems.

Type Description Agent Submission
ats_direct Direct application via ATS (e.g., Workday, Greenhouse, Lever) Varies
provider_hosted Provider controls the flow and delivers to ATS Full Support
platform_native Third-party platform owns the flow (e.g., Indeed Apply) Limited
email Legacy email-based application Not Supported
external_redirect Redirect to opaque external page Not Supported
custom Non-standard or conversational mechanisms Varies

The Application Workflow

OJCP defines a four-step process for an agent to move from discovery to application:

  1. Discover: The agent probes /.well-known/ojcp.json or queries the OJCP Registry to find providers.
  2. Search & Evaluate: The agent calls search_jobs using a CandidateContext for personalized ranking. Providers return a fit_score and fit_rationale to help the agent reason over the opportunity.
  3. Apply: The agent calls begin_application with an AgentDeclaration and a consent token.
  4. Track: The agent uses check_application_status to monitor the progress of the application.

Community Feedback and Considerations

Community discussion around the OJCP proposal has highlighted several points of contention and potential improvements:

"I find it offensive to use the word protocol when proposing a solution in the beginning at this early stage... glueing things together to make it work for your own use case isn't a worthy candidate to be a protocol."

Critics have raised concerns regarding the governance of the protocol, noting a lack of visible committee members beyond Recruitics. Others have questioned the impact of AI-driven applications on the recruiting industry, expressing concern that the process may devolve into "AI said this guy," while some users have suggested that the protocol could be extended to track company ghosting rates to protect job seekers.

Sources

Related

  • Dispatch
    Model Context Protocol (MCP) Roadmap Update

    The Model Context Protocol (MCP) is shifting toward agentic messaging, HTTP-native transport unification, and enterprise-grade agent identity to support autonomous cloud workloads.

  • Project
    tableau/tableau-mcp

    A suite of developer primitives using the Model Context Protocol (MCP) to integrate Tableau's data and analytics into AI applications.

  • Dispatch
    The Model Context Protocol: Overkill or the Future of Agentic Workflows?

    An exploration of the Model Context Protocol (MCP) versus traditional CLI tools, examining whether the proliferation of local processes is a necessary evolution for AI agents.

  • Dispatch
    Solving the MCP Onboarding Friction: The 'Hello Page' Pattern

    A technical look at how using HTTP Accept headers to provide human-readable guidance for Model Context Protocol (MCP) servers can drastically reduce support tickets and improve user onboarding.

  • Project
    apify/mcpc

    A universal CLI client for the Model Context Protocol (MCP) that enables AI agents and developers to interact with MCP servers via shell commands.