Monarchic LLC
ExplicitMem MCP
Auditable Memory
Benchmarks
ExplicitMem is a private memory MCP runtime with typed memory objects, graph links, bounded context synthesis, and receipt-backed retrieval. This page reports the current reproduced LongMemEval-S run beside published Mem0 and Supermemory website claims in a MemScore-style format, plus a live staging smoke for hosted MCP route latency.
Current ExplicitMem Run
Questions
500
Passed
499
Accuracy
99.8%
Faithfulness
100%
Retrieval recall@K
100%
Context tokens
796 avg
Local eval latency
30ms avg
Hosted write p50
3.45s
Hosted search p50
3.28s
Hosted context p50
3.48s
Retrieval receipts
100%
Protocol
synthesis
MemScore-Style Comparison
Quality / Latency / Context Tokens
| System | LongMemEval | Latency | Context Tokens | Summary | Source |
|---|---|---|---|---|---|
| ExplicitMem | 99.8% | 30ms local / 3.45s hosted write p50 | 796 avg | 99.8% / 30ms local / 796tok 499 of 500 LongMemEval-S questions passed under the synthesis answer protocol. Hosted staging MCP smoke measured 3.45s write p50, 3.28s search p50, and 3.48s context p50 across 10 successful samples. | Local reproduced run |
| Mem0 | 94.4% | 1.09s p50 | 6,787 avg | 94.4% / 1090ms / 6787tok Website-published LongMemEval result; not yet imported as a same-protocol provider artifact. | Published Mem0 research |
| Supermemory | 85.4% | <300ms | Not published | 85.4% / <300ms / n/a Website-published LongMemEval claim; context-token count was not published on the cited page. | Published Supermemory claim |
What We Reproduce
ExplicitMem's row is generated from a local 500-question LongMemEval-S benchmark artifact with reproduced accuracy, faithfulness, retrieval recall, latency, token count, and receipt coverage. Hosted latency is reported separately from a June 26, 2026 staging route smoke because it includes Auth0, the Monarchic API route, Lambda invocation, and MCP tool execution.
What We Cite
Mem0 and Supermemory values are public website claims. They are shown for market context and should not be read as same-harness reproduced provider results.
Next Gate
Same-protocol comparison requires provider artifacts with matching dataset, question order, judge, latency boundary, generated answers, context tokens, retrieved session ids, and hosted-route measurements.
Sources and Scope
- ExplicitMem artifact: internal reproduced benchmark report; hosted staging smoke on June 26, 2026 using `project_specific` scope, 10 successful samples, and the `https://staging-api.monarchic.io/mcp/explicitmem` route.
- Mem0 research page: mem0.ai/research and benchmark deep dive: AI Memory Benchmarks 2026
- Supermemory published claim: Supermemory vs Zep and MemScore format: MemoryBench MemScore
ExplicitMem results are benchmark-scoped and should not be presented as a general cross-dataset synthesis claim. Published third-party values are included only as cited market context until same-protocol provider artifacts are imported and validated. Hosted route latency is a staging infrastructure smoke, not the local LongMemEval harness latency.
Product Link
Use ExplicitMem
The product page translates these memory benchmark notes into the hosted MCP plan, bundle availability, and checkout path.