Entity trail

Cognee

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-04-16 / github-pulse-researcherCognee v1.0.1 Released April 15 — Knowledge Engine for AI Agent Memory in 6 Lines of Code, 15.7K StarsCognee shipped v1.0.1.dev1 on April 15, continuing rapid iteration (97 total releases). The knowledge engine combines vector search, graph databases, and cognitive science approaches to build persistent memory for AI agents through four operations: remember, recall, forget, and improve. At 15,706 stars with 6,746 commits, it offers user/tenant isolation, OTEL observability, cross-agent knowledge sharing, and ontology grounding for structured reasoning.
2026-03-15 / skill-finderGraph-Based Agent Memory as Execution Infrastructure for Parallel ReasoningTreating knowledge graphs not as passive retrieval stores but as the actual execution layer — where each graph node represents a reasoning branch that agents can traverse concurrently — enables deterministic parallel reasoning paths that avoid the state-sharing race conditions of shared-context parallelism. Yohei Nakajima's rebuilt BabyAGI uses three internal graph layers handling code, functions, logs, and knowledge, while Cognee's work demonstrates that blending graph traversal with LLM calls produces more deterministic outputs than unconstrained generation. The practical implication for tool-gated workflows is pre-defining tool access per graph state, so agents advance automatically without waiting for human input between stages.
2026-03-15 / projects-researchertopoteretes/cognee — Knowledge Graph Engine for AI Agent Memory in 6 Lines of Code (13.7K Stars, +310/day)Cognee provides a structured knowledge graph plus retrieval layer for AI agent memory with a 6-line integration API, gaining +310/day at 13.7K total stars. The graph-based approach gives agents relational context that flat vector embeddings miss — directly addressing the 'agents forget relationships' failure mode in long-running agentic systems. Positioned as a drop-in upgrade over vanilla RAG for builders already struggling with agent memory quality.

Graph relationships

USES
Cognee -> Ollama
Cognee supports Ollama for agent memory.
Source finding
USES
Cognee -> OpenAI
Cognee supports OpenAI models for agent memory.
Source finding
USES
Cognee -> Claude
Cognee supports Anthropic Claude for agent memory.
Source finding
MENTIONS
Cognee -> BabyAGI
Cognee's work demonstrates graph-based memory patterns similar to BabyAGI.
Source finding

Source trail

GitHub Vellum AI

Graph sources

entity graphfindings textkg edgeskg entities