Shared_Memory
Memory Framework based on the idea of Architectural Decision Records, shared knowledge between AI agents (Claude Code, Codex, Antigravity ...). Agent installable, vector (Postgres + pgvector) + graph (Neo4j) backbone, managed through a local gateway, supporting multiple local LLM backends, with full provenance, consolidation, telemetry and more
Warning. Serious findings were identified. Review the full report before connecting. Grades reflect the full trust pyramid: code, verification depth, and reputation. New projects cap at C until adoption is earned.
How we verified
Verified is a snapshot. Live keeps it current, and builds your track record.
⚡ Connect GitHub → continuous verification on every pushwhy connect →Who stands behind it
KanenasInGreece
Source: github_repo_search
Claim it to get a verified publisher badge, a free copy of our full audit findings, and direct contact for any high-priority issues we find. Or connect your repo for our deepest verification, Live Monitored: read-only, revoke anytime. What we access →
Install from
The grade above is for the source repository. Registries can serve a different version, so we mark the ones we were not able to read.
These names and descriptions are the publisher's own, read from the source code. We print them as written. Our assessment is the findings above, not this list.
hybrid_search_and_rerankSearch the shared memory: Tier-3 thematic/insight narratives for orientation, Tier-1 facts for precision, expanded through the entity graph.
save_artifactRequires a write-capable agent token: a read-only token receives an honest HTTP 403 role refusal from the gateway — expected, do not retry. Stores an artifact in shared memory via the Hive-Mind Gateway.
archive_reasoning_traceRequires a write-capable agent token: a read-only token receives an honest HTTP 403 role refusal from the gateway — expected, do not retry. Archive the agent's reasoning path as a memory record.
save_decisionRequires a write-capable agent token: a read-only token receives an honest HTTP 403 role refusal from the gateway — expected, do not retry. Save an architectural or design decision with full PROV-O provenance.
save_retrospectiveRequires a write-capable agent token: a read-only token receives an honest HTTP 403 role refusal from the gateway — expected, do not retry. Record an outcome for an existing Decision as a full retrospective record (own searchable record + Retrospective node behind the decision's HAD_OUTCOME trigger …
supersedeRequires a write-capable agent token: a read-only token receives an honest HTTP 403 role refusal from the gateway — expected, do not retry. Retract / supersede an existing FACT (decision 381/384). Decisions and retrospectives are refused with HTTP 400: supersession is the fact lifecycle. To overturn…
review_holdRequires a write-capable agent token: a read-only token receives an honest HTTP 403 role refusal from the gateway — expected, do not retry. Mark a summary's flagged stale source as reviewed-and-held (decision 384).
check_memory_healthFull-stack diagnostic for the shared-memory infrastructure.
memory_telemetryOperational telemetry snapshot from the gateway (GET /memory/telemetry).
record_lineageThis is a READ — GET /memory/status/{ref} makes no mutation, so it is on the gateway's read-role allowlist and a read-only agent token reaches it fine, no 403. "What happened to this record?" — its state, its dream-cycle stamps (applied → rem_reviewed → consolidated), and which summary it was folded…
graph_queryRun a READ-ONLY Cypher query against the knowledge graph.
VECTOR_SKILL_ENVthe token out of that config entirely by pointing at a client-only .envCOORDINATOR_URLAGENT_IDHEALTH_PROBE_TIMEOUT_SSEARCH_TIMEOUT_SSEARCH_TIMEOUT_FLOOR_SSEARCH_TIMEOUT_MAX_SSEARCH_TIMEOUT_FALLBACK_SSEARCH_SAFETY_FACTORSEARCH_OVERHEAD_SAGENT_TOKENexport by reading it out of a write-capable agent's skill .env from step 6 (theMEMORY_LOG_LEVELMEMORY_LOG_PATHPG_PASSWORDNEO4J_URINEO4J_USERNEO4J_PASSWORDPG_USERPG_HOSTPG_PORTPG_DATABASEPG_MAINTENANCE_DBGATEWAY_URLLLM_BACKENDS_JSONadds nothing to the host: an overnight of dreaming measured ~18,000 tokens, under a cent. With no LLM configured nothing dies — saves, search and the graph keep working; summaries and insights queue until a backend appearsNEO4J_MAX_POOLNEO4J_ACQUIRE_TIMEOUTLLM_MODELNREM_IDLE_THRESHOLD_SECNREM_ELIGIBILITY_RECHECK_SECNREM_POOL_PROBE_SECNREM_MAX_TOKENS_SUMMARYNREM_MAX_TOKENS_INSIGHTNREM_TRUNCATION_RETRY_FACTORNREM_FOLD_FAIL_WINDOWNREM_FOLD_FAIL_CAPNREM_INSIGHT_SLOT_INPUT_CHARSNREM_FORCED_SLOT_WAITBACKUP_ADVISORY_LOCK_KEYNREM_PRIORITY_ADVISORY_LOCK_KEYNREM_SWEEP_INTERVAL_SECNREM_TEMPERATUREDREAM_TEMPERATURE(load it text-only; set =0.6, Gemma degrades at lower temperatures) andCONSOLIDATION_RUNS_RETENTION_DAYSMOCK_LLMAGENT_ROLESdistinguishes one origin from another. Roles can narrow a token: =monitor:readENTITY_CONFUSABLE_SIMILARITYGATEWAY_LATENCY_WINDOWCLIENT_VERSIONS_MAXGATEWAY_AUDIT_LOG_PATHCONSOLIDATION_HEALTH_REFRESH_SECCONSOLIDATION_ORPHAN_TIMEOUT_SECTELEMETRY_CACHE_SENCODER_LATENCY_WINDOWPOOL_WAIT_WINDOWNEO4J_LATENCY_WINDOWENCODER_LATENCY_WARN_MSOUTBOX_AGE_WARN_STOKEN_VERIFY_WARN_PER_MINNREM_FOLD_ATTEMPT_WARNBREAKDOWN_AXIS_TOP_NDOMAIN_CONFUSABLE_SIMILARITYDREAM_METRICS_PATHLLM_CEILING_FLOORLLM_MIN_TOK_SEMBED_MIN_TOK_SEMBED_MAX_CONTEXT_TOKENSEMBED_CHARS_PER_TOKENEMBED_SAFETY_FACTOREMBED_TIMEOUT_FLOOR_SRERANK_MIN_CHARS_SRERANK_SAFETY_FACTORRERANK_TIMEOUT_FLOOR_SMAX_ENTITY_NAME_WORDSSLOT_AWARENVTOP_BINNVTOP_TIMEOUT_SECGPU_BUSY_PERCENTGPU_INDICESEMBEDDER_URLexpects, so points straight at it.RERANKER_URLSHIM_PORT, SHIM_TIMEOUT_S). Run it, point at it, restart the gateway. It bindsEMBED_RERANK_BUFFER_CAPLLM_DEFAULT_TARGETLLM_BACKENDSand it can be a pool: ="url@weight,…" load-balances the dreaming across severalLLM_FAIL_THRESHOLDLLM_FAIL_WINDOWLLM_COOLDOWNLLM_CHARS_PER_TOKEN_RATIOFIT_MARGINFIT_DEFAULT_OUTPUT_TOKENSLLM_MAX_INFLIGHT_WAIT_SLLM_MAX_INFLIGHT_POLL_SLLM_MAX_CAPACITY_WAITERSLLM_WEDGE_SUSPECT_AGELLM_USAGE_CAPTURE_CAP_BYTESLLM_AFFINITY_PREFIX_CHARSLLM_AFFINITY_TTLLLM_AFFINITY_MAX_INFLIGHTLLM_STEER_EXTRA_AGENT_NAMESCAPABILITY_PROBE_INTERVAL_SCAPABILITY_PROBE_RETRY_SCAPABILITY_PROBE_DOCSCAPABILITY_PROBE_DOC_CHARSCAPACITY_NEO4J_FALLBACK_BYTESCAPACITY_PG_MEM_ALLOWANCE_BYTESCAPACITY_EMBEDDER_MEM_ALLOWANCE_BYTESCAPACITY_GATEWAY_MEM_ALLOWANCE_BYTESCAPACITY_OS_MEM_MARGIN_BYTESNEO4J_HEAP_MAXNEO4J_PAGECACHESet the small-host preset (NEO4J_HEAP_INITIAL/MAX, NEO4J_PAGECACHE) from .env.example; preflight's RAM check tells you which tier you are on.CPU_ENCODER_REPLICASand =0, and what you run never diverges from what ships. If you have oneGPU_ENCODER_REPLICASIntel, AMD and NVIDIA), off by default — the choice is two .env lines, =1HEALTH_CACHE_TTL_SEMBED_MAX_CHARSALLOW_UNAUTHENTICATED_PROVIDER_KEYSTOKEN_LIFECYCLE_SUM_INTERVAL_SXDG_RUNTIME_DIRPROXY_BINDThe gateway binds to localhost by default. =0.0.0.0 opts into wider exposure — doGATEWAY_UDS_PATHGATEWAY_UDS_MODESECURE_ENV_FILESHARED_MEMORY_PROJECTCOORDINATOR_UDSSM_PRE_UPDATE_VERSIONGATEWAY_UNITEMBED_URLPROJECT_ALIASESSMEM_ONTOLOGY_PATHMIN_ENTITY_NAME_LENPOOL_STATUS_URLPROJECT_CONFUSABLE_SIMILARITYAUDIT_LOG_PATHWRITE_QUIESCE_SECREM_SUMMARY_THRESHOLDREM_TEMPERATUREREM_MAX_TOKENS_SOLOREM_MAX_TOKENS_PER_FACTREM_MAX_TOKENS_PER_SUMMARYREM_TRUNCATION_RETRY_FACTORREM_TRUNCATION_SPECIMEN_CHARSREM_MAX_ATTEMPTSREM_STARVED_THRESHOLDSHIM_VLLM_URLand refuses any path it was not built to map. Four env settings (, SHIM_HOST,SHIM_HOSTand refuses any path it was not built to map. Four env settings (SHIM_VLLM_URL, ,SHIM_PORT, SHIM_TIMEOUT_S). Run it, point RERANKER_URL at it, restart the gateway. It bindsSHIM_TIMEOUT_SSHIM_PORT, ). Run it, point RERANKER_URL at it, restart the gateway. It bindsSECURE_ENV_SECRET_FILE_MAX_BYTESCREDENTIALS_DIRECTORYagent_tokensPROJECT_CLOSE_MATCH_CUTOFFPROJECT_ROOTSTool annotations
No tools have read-only/destructive annotations
Add readOnlyHint or destructiveHint annotations to every tool so hosts can warn users before invoking.
All four hints declared on every tool
11/11 tools missing one or more hints — hybrid_search_and_rerank (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint); save_artifact (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint); archive_reasoning_trace (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint), +8 more. OpenAI's directory rejects tools where any of the four hints are missing or non-boolean.
For every tool, set all four hints (readOnlyHint, destructiveHint, idempotentHint, openWorldHint) to explicit true/false values that match the handler’s actual behaviour.
No access to sensitive paths
Reads sensitive paths: ~/.claude/skills/shared-memory/.env, ~/.codex/skills/shared-memory/.env, ~/.gemini/skills/shared-memory/.env
Remove reads of sensitive system paths. If you genuinely need them, document why in the README.
Claim the listing to review these findings one by one and send us a correction where you disagree, straight to the team. Claiming also means we tell you when the grade moves, and reach you first if we find anything urgent.
[](https://m8ven.ai/mcp/kanenasingreece/shared_memory)?variant=verified from the URL.Vetting this one by hand? Tool Check is an MCP that scores other MCPs. Add it once and ask Claude, ChatGPT, or any MCP client to grade a server, surface CVEs, check the publisher, and suggest safer alternatives — before you install.
https://m8ven.ai/api/mcp/tool-check