Demo: How to Build Your Own AI Memory for Agents
Blog post from Memgraph
Memgraph Context Graph is an open-source AI memory layer in the Memgraph AI Toolkit that captures coding-agent sessions, tool calls, prompts, outputs, resources, and actions in a shared graph so agents such as Claude Code and Codex can recall work across sessions and providers. It uses hooks and specialized connectors to record raw activity, while an optional reconciliation pass extracts entities and facts into a knowledge graph that helps locate relevant source events without requiring LLM calls by default. In a live demonstration, Claude Code and Codex were connected to the same Memgraph-backed memory, allowing Codex to report on Claude’s completed work and an active GitHub issue in progress. The system supports configurable models, connectors, token-limited recall, user-specific and shared team spaces, secret redaction, and selective sharing, though storage metrics and broader harness support remain areas for development. Early benchmark results reported 87.4% accuracy on LongMemEval-S and 63.1% on BEAM 100K, with future work focused on improving scale, accuracy, latency, and token efficiency.
| Trend | Post Mentions | Total Month Mentions | Posts | Companies | MoM |
|---|---|---|---|---|---|
| LLM | 3 | 406 | 98 | 57 | -94% |
| Vector Search | 2 | 47 | 27 | 22 | -98% |
| AI Agents | 1 | 1,056 | 148 | 77 | -85% |
| Developer Experience | 1 | 26 | 17 | 8 | -95% |
| Real-time | 1 | 386 | 100 | 56 | -91% |
| Secrets Management | 1 | 304 | 76 | 38 | -90% |
Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.