Radar calibration
Scoring and Role Derivation
All scores on 0–5 scale. Dimensions: CM = Centroid Match, DF = Differentiation, HE = Human-Edge, AD = Adoption Depth, CI = Claim/Implementation Integrity. Evidence snapshot: 2026-05-10.
Denominator: full corpus, N=32, with in-cluster N=20 separated from adjacent, reference, and edge-case rows. Source: Manual-audit/CALIBRATION-TABLE-RADAR.md. Reader meaning: calibration shows the evidence trail behind public role labels and sector-health metrics.
Decision Rules
Cluster roles are derived algorithmically. Rules are applied in order; first match wins.
| Priority | Condition | Role |
|---|
| 1 | Human-Edge ≤ 1 | Inactive Scaffold |
| 2 | HE ≥ 4 AND AD ≥ 4 AND CM ≥ 3 | Category Leader |
| 3 | CM ≤ 2 AND HE ≥ 3 | Adjacent Infrastructure |
| 4 | DF ≥ 3 AND HE ≥ 2 | Differentiated Niche |
| 5 | CM ≥ 4 AND DF ≤ 2 AND HE ≥ 2 | Strong Default |
| 6 | Remaining | Baseline Implementation |
Score Context
| Dimension | Scale | Reader Meaning | Source |
|---|
| Centroid Match | 0-5 | Similarity to the accepted MCP memory-server centroid. | Centroid traits and calibration rows. |
| Differentiation | 0-5 | Observable distinction from the default sector pattern. | Calibration derivation notes. |
| Human-Edge | 0-5 | Sustained engineering evidence visible at the snapshot date. | Audit traces and repository evidence. |
| Adoption Depth | 0-5 | Observed adoption signal at the snapshot date. | Repository metadata and calibration notes. |
| Claim/Implementation Integrity | 0-5 | Fit between public claims and implementation evidence. | README/package/source trace checks. |
In-Cluster MCP Memory Servers (20)
| # | Repo | CM | DF | HE | AD | CI | Cluster Role |
|---|
| 1 | thomasahle/claude-mem | 5 | 4 | 4 | 5 | 5 | Category Leader |
| 2 | mcp-memory-keeper | 5 | 1 | 2 | 2 | 4 | Strong Default |
| 3 | ultratool/context-portal (ConPort) | 4 | 3 | 4 | 3 | 5 | Differentiated Niche |
| 4 | datawiz168/mempalace | 5 | 5 | 3 | 2 | 4 | Differentiated Niche |
| 5 | 0xAmey/continuum | 5 | 2 | 3 | 1 | 4 | Strong Default |
| 6 | nikhiltidke/contextgo | 5 | 4 | 3 | 1 | 4 | Differentiated Niche |
| 7 | mem0ai/mem0-mcp | 4 | 0 | 1 | 1 | 1 | Inactive Scaffold |
| 8 | contextfs | 4 | 0 | 1 | 0 | 2 | Inactive Scaffold |
| 9 | persistent-context-mcp | 5 | 0 | 0 | 0 | 0 | Inactive Scaffold |
| 10 | campfirein/byterover-cli | 3 | 5 | 5 | 2 | 4 | Differentiated Niche |
| 11 | Gentleman-Programming/engram | 5 | 4 | 5 | 3 | 4 | Differentiated Niche |
| 12 | doobidoo/mcp-memory-service | 4 | 5 | 5 | 5 | 4 | Category Leader |
| 13 | alioshr/memory-bank-mcp | 3 | 1 | 2 | 5 | 5 | Baseline Implementation |
| 14 | 24kchengYe/MemoMind | 4 | 5 | 4 | 2 | 4 | Differentiated Niche |
| 15 | Ori-Mnemos | 5 | 5 | 4 | 1 | 5 | Differentiated Niche |
| 16 | forgetful | 5 | 5 | 4 | 2 | 5 | Differentiated Niche |
| 17 | ogham-mcp | 4 | 5 | 3 | 3 | 5 | Differentiated Niche |
| 18 | dazeb/cline-mcp-memory-bank | 4 | 1 | 0 | 1 | 4 | Inactive Scaffold |
| 19 | IzumiSy/mcp-duckdb-memory-server | 3 | 2 | 2 | 1 | 5 | Baseline Implementation |
| 20 | maydali28/memcp | 5 | 4 | 1 | 0 | 1 | Inactive Scaffold |
Adjacent Infrastructure (8)
| # | Repo | CM | DF | HE | AD | CI | Cluster Role |
|---|
| 21 | cognee-ai/cognee | 2 | 5 | 4 | 3 | 4 | Adjacent Infrastructure |
| 22 | mem0ai/mem0 | 1 | 5 | 5 | 5 | 5 | Adjacent Infrastructure |
| 23 | zep-ai/graphiti | 0 | 4 | 5 | 5 | 5 | Adjacent Infrastructure |
| 24 | letta-ai/letta | 0 | 4 | 5 | 4 | 5 | Adjacent Infrastructure |
| 25 | NevaMind-AI/memU | 2 | 5 | 5 | 4 | 4 | Adjacent Infrastructure |
| 26 | aiming-lab/SimpleMem | 2 | 4 | 4 | 4 | 3 | Adjacent Infrastructure |
| 27 | sanonone/kektordb | 2 | 5 | 2 | 1 | 5 | Adjacent Infrastructure* |
| 28 | MemTensor/MemOS | 2 | 5 | 5 | 4 | 4 | Adjacent Infrastructure |
Reference / Self (2)
| # | Repo | CM | DF | HE | AD | CI | Cluster Role |
|---|
| 29 | Skitchy/rekindle (Author's project) | 5 | 5 | 5 | 1 | 5 | Differentiated Niche† |
| 30 | modelcontextprotocol/servers (memory) | 3 | 2 | 4 | 5 | 5 | Reference Implementation‡ |
Out-of-Scope Edge Cases (2)
| # | Repo | CM | DF | HE | AD | CI | Cluster Role |
|---|
| 31 | supermemoryai/supermemory-mcp | 3 | 2 | 2 | 3 | 3 | Baseline Implementation |
| 32 | coleam00/claude-memory-compiler | 2 | 3 | 2 | 2 | 2 | Edge Case§ |
Boundary and Override Notes
* kektordb: Algorithm assigns Differentiated Niche (DF 5 ≥ 3, HE 2 ≥ 2). Overridden to Adjacent Infrastructure. Product is an embeddable vector database with an MCP flag — not an MCP memory server. High differentiation reflects category difference, not innovation within the cluster. Centroid match (2/9) confirms low cluster membership.
† Rekindle: Author's project. COI disclosed per SC-010 protocol. Scored on same rubric as all other repos. No score adjustments. Algorithm assigns Differentiated Niche (DF 5 ≥ 3, HE 5 ≥ 2). Role stands.
‡ modelcontextprotocol/servers: Algorithm assigns Category Leader (HE 4 ≥ 4, AD 5 ≥ 4, CM 3 ≥ 3). Classified as Reference Implementation. This is the official MCP protocol standard maintained by Anthropic — the benchmark against which the sector is built, not a competitor within it.
§ claude-memory-compiler: Algorithm assigns Differentiated Niche (DF 3 ≥ 3, HE 2 ≥ 2). Classified as Edge Case. Product is a knowledge compilation pipeline via Claude Code hooks — not an MCP server. Centroid match (2/9) confirms fundamental category difference.
Cluster Composition
In-Cluster Only (n=20)
| Role | Count | % | Repos |
|---|
| Category Leader | 2 | 10% | claude-mem, mcp-memory-service |
| Strong Default | 2 | 10% | mcp-memory-keeper, Continuum |
| Differentiated Niche | 9 | 45% | ConPort, MemPalace, ContextGO, ByteRover, Engram, MemoMind, Ori-Mnemos, forgetful, ogham-mcp |
| Baseline Implementation | 2 | 10% | memory-bank-mcp, mcp-duckdb-memory-server |
| Inactive Scaffold | 5 | 25% | mem0-mcp, contextfs, persistent-context-mcp, cline-mcp-memory-bank, memcp |
Full Corpus (n=32)
| Role | Count | % | Repos |
|---|
| Category Leader | 2 | 6% | claude-mem, mcp-memory-service |
| Strong Default | 2 | 6% | mcp-memory-keeper, Continuum |
| Differentiated Niche | 10 | 31% | ConPort, MemPalace, ContextGO, ByteRover, Engram, MemoMind, Ori-Mnemos, forgetful, ogham-mcp, Rekindle† |
| Baseline Implementation | 3 | 9% | memory-bank-mcp, mcp-duckdb-memory-server, supermemory-mcp |
| Inactive Scaffold | 5 | 16% | mem0-mcp, contextfs, persistent-context-mcp, cline-mcp-memory-bank, memcp |
| Adjacent Infrastructure | 8 | 25% | cognee, Mem0, Graphiti, Letta, memU, SimpleMem, kektordb*, MemOS |
| Reference Implementation | 1 | 3% | mcp-server (official)‡ |
| Edge Case | 1 | 3% | claude-memory-compiler§ |
Sector Health Metrics
In-Cluster (n=20)
| Metric | Value | Interpretation |
|---|
| Convergence Index | 4.35 / 5 | High — strong surface convergence within the MCP memory server cluster |
| Differentiation Index | 3.1 / 5 | Moderate — significant innovation exists alongside convergence |
| Viability Rate | 60% (12/20) | Moderate — majority of projects have real engineering |
| Scaffold Rate | 25% (5/20) | Elevated — 1 in 4 in-cluster projects is minimally developed |
| Integrity Index | 3.8 / 5 | Healthy — most projects deliver what they claim |
| Adoption Concentration | 15% (3/20) | Low — adoption is concentrated in a few projects |
Full Corpus (n=32)
| Metric | Value | Interpretation |
|---|
| Convergence Index | 3.5 / 5 | Moderate — adjacent infrastructure pulls average down |
| Differentiation Index | 3.4 / 5 | Moderate-High — broad innovation across the ecosystem |
| Viability Rate | 66% (21/32) | Healthy — two-thirds of the ecosystem has real engineering |
| Scaffold Rate | 16% (5/32) | Moderate — scaffolds are diluted by adjacent infrastructure |
| Category Leakage | 25% (8/32) | Significant — "memory" keyword pulls in adjacent categories |
| Integrity Index | 3.9 / 5 | Healthy — ecosystem-wide integrity is strong |
Source Artifacts
Scores, roles, health metrics, and boundary notes are derived from accepted MCP Memory Servers artifacts. Found a stale repo, wrong source trace, or boundary classification issue? Send a correction to hello@diversum.dev with the repo name, displayed value, and replacement source.
| Artifact | Source Role | Visible Metric Families |
|---|
| Manual-audit/CALIBRATION-TABLE-RADAR.md | Score and role derivation | All calibration rows, role rules, and boundary notes. |
| Manual-audit/CENTROID-MCP-MEMORY-32.md | Centroid denominator and trait counts | Centroid Match interpretation and in-cluster boundary. |
| Reports/VISUAL-DATA.md | Chart-ready aggregations | Composition and sector-health values. |
| Reports/SECTOR-RADAR-MCP-MEMORY-SERVERS.md | Public narrative report | Interpretation, limitations, and correction path. |
Corrections
If a score, role, or boundary classification appears wrong, contact hello@diversum.dev with the repo name, disputed row, and source evidence for review.