How well can AI be an MCP server or client?
No SDK — just the raw model.
2025-11-25
57% 73%
no docs · full docs
2026-07-28
5% 76%
no docs · full docs
Client
38 scenarios · 516 trials · $170.95
2025-11-25
49% no docs · 64% full docs
no docs full docs
| Model | No docs | Full docs | 0 → 100% |
|---|---|---|---|
| openai/gpt-5.6-sol | 62% | 81% | |
| anthropic/claude-opus-4-8 | 57% | 67% | |
| openai/gpt-5.6-luna | 52% | 62% | |
| anthropic/claude-sonnet-5 | 52% | 57% | |
| openai/gpt-5.6-terra | 48% | 62% | |
| moonshotai/kimi-k2.7-code | 24% | 57% |
| Capability | No docs | Full docs |
|---|---|---|
| Tools, resources & prompts | 79% | 79% |
| Streaming & transport | 73% | 83% |
| Lifecycle & session | 53% | 67% |
| Authorization & identity | 13% | 47% |
| Elicitation | 8% | 25% |
2026-07-28
7% no docs · 72% full docs
no docs full docs
| Model | No docs | Full docs | 0 → 100% |
|---|---|---|---|
| openai/gpt-5.6-sol | 18% | 86% | |
| anthropic/claude-opus-4-8 | 14% | 64% | |
| openai/gpt-5.6-terra | 5% | 82% | |
| anthropic/claude-sonnet-5 | 5% | 64% | |
| openai/gpt-5.6-luna | 0% | 77% | |
| moonshotai/kimi-k2.7-code | 0% | 61% |
| Capability | No docs | Full docs |
|---|---|---|
| Authorization & identity | 15% | 47% |
| Multi-round tool results | 8% | 73% |
| Request-header contract | 0% | 100% |
| Lifecycle & session | 0% | 92% |
| Stateless mode | 0% | 92% |
| Async tasks | 0% | 88% |
Server
16 scenarios · 228 trials · $72.38
2025-11-25
73% no docs · 93% full docs
no docs full docs
| Model | No docs | Full docs | 0 → 100% |
|---|---|---|---|
| openai/gpt-5.6-luna | 90% | 90% | |
| anthropic/claude-sonnet-5 | 80% | 90% | |
| anthropic/claude-opus-4-8 | 75% | 100% | |
| openai/gpt-5.6-terra | 75% | 90% | |
| openai/gpt-5.6-sol | 65% | 85% | |
| moonshotai/kimi-k2.7-code | 55% | 100% |
| Capability | No docs | Full docs |
|---|---|---|
| Tools, resources & prompts | 83% | 100% |
| Streaming & transport | 83% | 94% |
| Errors & origin checks | 75% | 100% |
| Lifecycle & session | 67% | 96% |
| Elicitation | 58% | 75% |
2026-07-28
0% no docs · 84% full docs
no docs full docs
| Model | No docs | Full docs | 0 → 100% |
|---|---|---|---|
| openai/gpt-5.6-sol | 0% | 100% | |
| anthropic/claude-opus-4-8 | 0% | 94% | |
| openai/gpt-5.6-luna | 0% | 89% | |
| anthropic/claude-sonnet-5 | 0% | 83% | |
| moonshotai/kimi-k2.7-code | 0% | 78% | |
| openai/gpt-5.6-terra | 0% | 61% |
| Capability | No docs | Full docs |
|---|---|---|
| Streaming & transport | 0% | 100% |
| Async tasks | 0% | 100% |
| Elicitation | 0% | 83% |
| Request-header contract | 0% | 83% |
| Lifecycle & session | 0% | 79% |
| Stateless mode | 0% | 75% |