Contents

  1. The "MCP-Ready" Checkbox Problem
  2. KanseiLink's 3-Tier Agent Readiness Framework
  3. The verified Tier: A Tight, High-Trust Club
  4. The connectable Tier: 13%–100% Wild Range
  5. Grade × Readiness Matrix: The Surprising Misalignments
  6. Enterprise Procurement Guide: 5 Questions That Actually Matter
  7. For SaaS Vendors: How to Earn verified Status
  8. FAQ

The "MCP-Ready" Checkbox Problem

As AI agent adoption accelerates through 2026, "MCP-ready" is becoming a new line item in enterprise SaaS procurement evaluation. Vendors announce MCP server availability in press releases; procurement teams check the box and move on to the next criterion.

KanseiLink's real-world Agent Readiness data from 225+ services shows this approach is fundamentally broken.

"Are you MCP-ready?" is the wrong question. The right question is: "What Agent Readiness tier are you, and what is your actual success rate?"

KanseiLink Agent Readiness Data Summary (April 2026)

89%
Average success rate
across verified tier
13–100%
Success rate range
within connectable tier
6
Services holding
verified status
87%
Of "MCP-ready" services
that are connectable or below

KanseiLink's 3-Tier Agent Readiness Framework

KanseiLink evaluates MCP service Agent Readiness across 3 tiers. This classification is independent from the AEO grade (AAA/AA/A/BBB etc.) — they measure different dimensions.

🟢 verified — Battle-Tested

Sufficient real-world agent call data has accumulated and success rate ≥80% is confirmed. Major error patterns and known workarounds are documented. Recommended for production agent workflow integration.

KanseiLink April 2026: 6 services | Success rate range: 83–94%
🟡 connectable — Technically Reachable, Unproven in Production

An official MCP server or sufficient API exists and is technically connectable, but real-world agent data is insufficient or success rate falls below the verified threshold. "Can connect" and "agents can use it reliably" are two different things.

KanseiLink April 2026: majority of MCP services | Success rate range: 13–100% (heavily dependent on data volume)
info_only — Reference Only

No MCP server exists; API access is limited or absent. KanseiLink can provide basic service information (overview, category, pricing tier) but the service is not operable by agents.

Services in evaluation or with MCP on the product roadmap
Important: verified Is Not Static

Agent Readiness tier changes over time. Improving success rates and accumulating data can promote a service from connectable to verified. Conversely, API breaking changes or declining success rates can demote a service from verified to connectable. KanseiLink data is continuously updated.

The verified Tier: A Tight, High-Trust Club

As of April 2026, KanseiLink's verified tier contains 6 services — every one of them carrying an AAA AEO grade. Their success rates show a striking consistency.

Service AEO Grade Success Rate Sample (n) Avg Latency Agent Readiness
Shopify Japan AAA 94% n=50+ 🟢 verified
Money Forward Cloud AAA 93% n=50+ 🟢 verified
Slack AAA 91% n=113 163ms 🟢 verified
freee AAA 90% n=98+ 🟢 verified
Backlog AAA 90% n=50+ 🟢 verified
Notion AAA 83% n=50+ 🟢 verified

The verified tier averages 89% success. The floor is 83% (Notion), ceiling 94% (Shopify Japan). Low variance, high predictability — the profile you need to build production agent workflows against.

What verified Services Have in Common

All 6 hold AAA AEO grade, publish official MCP servers as npm packages, have accumulated substantial real-world usage data, and invest in documenting their known error patterns and workarounds. The formula: official MCP server × sustained agent usage × quality management.

The connectable Tier: 13%–100% Wild Range

connectable means "official MCP exists but unproven" — but the distribution within that tier is anything but uniform. KanseiLink data reveals a shocking spread.

Service AEO Grade Success Rate Sample (n) Top Errors Agent Readiness
Microsoft Teams A 100% small 🟡 connectable
LINE Messaging A 100% small 🟡 connectable
PostgreSQL MCP A 100% small 🟡 connectable
kintone AA 79% n=50+ api_error, invalid_input 🟡 connectable
Asana AA 67% moderate api_error 🟡 connectable
Garoon AA 67% moderate api_error 🟡 connectable
Chatwork AA 66% n=123 api_error (24x), search_miss (10x) 🟡 connectable
Sansan AA 61% n=36 api_error (9x), search_miss (5x) 🟡 connectable
Zapier A 13% n=9 search_miss (7x) 🟡 connectable

Within the same connectable tier, Microsoft Teams and Zapier are separated by 87 percentage points. Treating connectable as a uniform category means treating a 13% failure rate and a 100% early-data service identically.

⚠️ Caution on 100% connectable Services

Microsoft Teams, LINE Messaging, and PostgreSQL MCP show 100% success rates — but this likely reflects limited data points rather than proven at-scale reliability. Small samples can achieve 100% before edge cases emerge at volume. This is precisely why they remain connectable rather than verified: the data hasn't accumulated to confirm sustained production performance.

Grade × Readiness Matrix: The Surprising Misalignments

Mapping AEO grade against Agent Readiness tier surfaces an important insight: AEO grade and Agent Readiness tier measure different things.

AEO grade (AAA/AA/A/BBB) reflects API design quality, documentation depth, MCP server implementation completeness, and security posture — a design-time and architecture quality score. Agent Readiness tier reflects what actually happens when agents call the service at runtime — an operational performance score.

A-grade Zapier is connectable with 13% success — design quality and operational reality diverge. A-grade Microsoft Teams is connectable with 100% in early data — high design quality, insufficient data for verification. The grade tells you about the blueprint; the Readiness tier tells you about the building.

Practical Procurement Interpretation

AAA + verified = Production recommended. Safe for core enterprise agent workflows.
AA/A + connectable (high-n, high success) = Adopt cautiously. Pilot before production rollout.
AA/A + connectable (low-n or low success) = Pilot only. Request AEO improvement commitment from vendor.
A + connectable (Zapier-type) = Not recommended for production at this time.

Enterprise Procurement Guide: 5 Questions That Actually Matter

Retire "Are you MCP-ready?" and replace it with these five questions.

  1. What is your Agent Readiness tier? (verified / connectable / info_only)
    Check KanseiLink's AEO evaluation report or ask the vendor directly. An "MCP-ready" answer confirms connectable or above — but that alone is insufficient for a production commitment.
  2. What is your actual success rate, and what is the sample size n?
    n=9 at 100% and n=113 at 91% represent completely different confidence levels. Target n≥50 with ≥80% success as minimum thresholds for production integration.
  3. Are your major error patterns and workarounds documented?
    verified services have known issues and remediation steps documented. Absence of this documentation means high incident response cost when things break in production.
  4. What are your success rate and reliability trends over the last 30 days?
    API spec changes and feature launches can cause sudden success rate drops. Directional momentum matters as much as the current figure.
  5. Does the vendor commit to continuous AEO score improvement?
    Is the vendor actively improving their verified status? KanseiLink registration and ongoing data contribution is one measurable indicator of that commitment.

For SaaS Vendors: How to Earn verified Status

From 2026 onward, enterprise procurement teams will evaluate both AEO grade and Agent Readiness tier. "MCP-ready" alone no longer differentiates. The race to verified has begun.

The path to verified:

Is your service on the path to verified?

KanseiLink's AEO audit maps your current Agent Readiness tier and success rate against the verified threshold — and builds the improvement roadmap to get there.

Request an AEO Audit

FAQ

What is the difference between verified and connectable?

verified means success rate ≥80% with sufficient accumulated data (generally n≥50 equivalent). connectable means an official MCP server exists but real-world data is insufficient or success rate falls below the threshold. April 2026: 6 verified services cluster at 83–94% success; connectable spans 13–100%.

Can I trust a vendor that claims "MCP-ready"?

Not without asking for the tier and success rate. Zapier is A-grade connectable at 13%; Microsoft Teams is A-grade connectable at 100% (small sample). The "MCP-ready" label spans both. Always ask for the success rate and n.

How does a service achieve verified status?

Requirements: (1) success rate ≥80%, (2) sufficient data accumulation (n≥50 equivalent), (3) documented error patterns and workarounds. Practically: publish an official MCP server, build real agent usage, engage KanseiLink's AEO improvement program to accelerate data collection.

Data Disclosure

Data cited in this article is from KanseiLink MCP system agent call logs. Sample sizes vary significantly by service; small-sample services (Teams, LINE Messaging, etc.) may show volatile figures that stabilize with more data. Agent Readiness tiers are updated regularly and current ratings may differ from those at time of publication. Verify current ratings via the KanseiLink MCP server.