HR SaaS AEO Map 2026

HR SaaS AEO Landscape — How It Differs from Accounting

While accounting SaaS has emerged as AEO's frontrunner category, HR SaaS remains in an earlier, more turbulent phase of agent integration. The sensitivity of the data — employee records, attendance logs, payroll details — drives more complex API scope design and OAuth implementation, raising the bar for agent-compatible integrations.

Of the eight HR services tracked by KanseiLink, only freee HR offers an official MCP server (via the freee integrated MCP). The other seven are API-only, forcing agents to implement direct REST integrations with varying levels of documentation quality. This structural gap is the root cause of the category's low average success rates.

SmartHR — BBB Grade: The 39% Problem

SmartHR is the leading cloud HR platform in Japan — used by companies ranging from listed enterprises to growing SMBs for employee records, contracts, and year-end tax adjustments. It exposes a RESTful API for developers, and agent integration is theoretically well-supported. The reality of the data tells a different story.

Across 89 agent runs, success rate is 39% — one of the lowest among all major services tracked by KanseiLink. Seven distinct agents have attempted integration, amplifying confidence in the dataset. Three error categories account for nearly all failures:

SmartHR error analysis

1. API errors (36 incidents — highest count)
The majority trace to agents calling deprecated v1 endpoints. SmartHR has migrated its API to v2, but agents default to v1 without explicit instruction, generating a cascade of errors.

2. Auth errors and expiry (11 combined)
OAuth scope documentation is incomplete. Agents cannot determine required scopes in advance and must guess — a process that generates auth failures and requires manual developer intervention. The confirmed workaround: pre-request employees.read, attendance.read, and payroll.read before the auth flow.

3. Search misses (7 incidents)
Semantic queries like "manage employee onboarding" or "issue withholding tax certificate" consistently fail to surface SmartHR in search results. A search for "源泉徴収票を発行したい" (issue withholding tax certificate) returned freee, Money Forward, and KING OF TIME instead. SmartHR's semantic discoverability is a meaningful gap.

freee HR — A Grade: Official MCP, High Potential

freee HR has no live agent run data to report, so success rate and latency figures are not available. However, freee's integrated MCP server covers the HR domain — employee data, attendance, payroll, and year-end adjustment — under the same server that powers freee accounting. Structural readiness is solid.

Projecting from freee's accounting domain telemetry (90.3% success, 206 runs) provides reasonable grounds for optimism: the same MCP server, the same authentication plumbing, the same developer standards. For builders constructing HR AI assistants or automation agents, freee HR is the lowest-risk option in the category. That said, a pilot testing period before production deployment is recommended until live HR-domain telemetry accumulates.

KING OF TIME — BBB Grade: Attendance API in Practice

KING OF TIME is Japan's market-leading attendance management system by install base. Its REST API uses API key authentication and supports time-punch data retrieval, shift management, and overtime calculation. No official MCP server exists, but the relatively simple API design contributes to a 65% success rate — far ahead of SmartHR's 39%.

All 12 errors in the dataset are classified as API errors, with zero authentication failures. API key auth, simpler than OAuth2, reduces agent-side authentication failure risk. The tradeoff is key management responsibility — API key rotation and storage must be handled by the agent implementation layer.

Average latency of 199ms is meaningfully better than SmartHR (301ms), making it suitable for batch attendance data collection workflows. An official MCP server has not been announced, but the clean API design makes it a prime candidate for a community-built MCP wrapper.

Three-Way Comparison and Implementation Recommendations

SmartHR Integration Guide for Agent Developers

For teams building AI agents on top of an existing SmartHR deployment, the following steps will substantially improve success rates:

1. Use v2 endpoints — Target https://api.smarthr.jp/api/v2/ explicitly. Most failures in the dataset trace to v1 usage. v1 should be treated as deprecated.
2. Pre-request OAuth scopes — Explicitly include employees.read, attendance.read, and payroll.read in the authorization request before any API calls.
3. Review scope documentation manually — Check the SmartHR developer docs (developer.smarthr.jp) for the current scope list before implementing. Do not rely on agent inference alone.
4. Set generous timeouts — With average latency at 301ms and tail latencies higher, set minimum timeouts of 5–10 seconds to avoid false failures.
5. Implement retry logic — Given 36 API errors in the dataset, exponential backoff retry is essential for production reliability.
6. Use explicit service names in queries — Saying "SmartHR" explicitly in search queries avoids semantic search misses. Do not rely on intent-only queries when targeting SmartHR.

The broader story here is one of category divergence. Accounting SaaS has established a clear AEO benchmark; HR SaaS is still finding its footing. H2 2026 is shaping up to be the pivotal period for HR platform MCP adoption. The category standard is unset — and the first HR platform to ship a polished official MCP server could meaningfully reshape the competitive landscape.