SynapseGrid
Labs
View GitHub
BRAINCORE

Autonomous memory for long-running AI infrastructure.

BrainCore preserves operational knowledge as evidence-backed context so AI agents can retrieve facts, incidents, and decisions instead of rediscovering them in every session. The public production corpus reports 26,966 extracted facts, 9,074 tracked entities, and P50 retrieval at 71.6 ms.

BrainCore memory system social banner
WHAT_IS_BRAINCORE

BrainCore turns operational history into retrievable agent memory.

BrainCore is a public SynapseGrid Labs project for memory-first AI infrastructure. It archives incidents and sessions, extracts evidence-backed facts, and exposes temporal knowledge for automation systems that need continuity.

Evidence-backed context

Operational facts are tied to source evidence so agents can inspect where memory came from.

Fast memory retrieval

Production-corpus retrieval is measured at P50 71.6 ms and P95 85.2 ms.

Release-aware public surface

Public claims and metrics are tied to a claims binder and production benchmark evidence.

PUBLIC_METRICS

Production memory metrics.

26,966

Preserved facts

Structured operational memory for AI infrastructure and agent workflows.

9,074

Tracked entities

Entity-aware context helps agents connect systems, projects, incidents, and decisions.

623

Published memories

Condensed memory artifacts turn operational history into reusable context.

71.6

P50 retrieval ms

Fast retrieval targets keep memory usable inside agent work loops.

85.2

P95 retrieval ms

Tail latency stays visible because memory is infrastructure, not decorative context.

14,061

Evidence segments

Evidence segments support traceable retrieval and source-backed public claims.

POSITIONING

Operational memory, not generic chat memory.

BrainCore is positioned for AI infrastructure and agent systems: deterministic parsers, trust classes, temporal facts, PostgreSQL/pgvector retrieval, provenance, and memory MCP tools. It should not be described as hosted SaaS or universal chat memory.

release v1.1.6

stack TypeScript + Python + PostgreSQL + pgvector

retrieval P50 71.6 ms / P95 85.2 ms

proof claims binder + benchmark evidence

RELATED_TOPICS

BrainCore search context.