Nocturne Memory
A lightweight, rollbackable, and visual Long-Term Memory Server for MCP Agents. Say goodbye to Vector RAG and amnesia. Empower your AI with persistent, graph-li…
Horon is described as an executable memory and cognitive harness for AI agents. It is intended to let an agent learn, remember, and build its own guardrails. Its adaptive memory learns the paths along which an AI actually reads its memory, so related memories become easier to surface next time.
It can block a tool before execution when preconditions are not met and tell the AI what step is still missing. It supports Claude Code, Codex, Antigravity, OpenCode, and DeepSeek Harness. A visual console is provided for viewing the memory graph, the circuits currently powered, and the associations between memories.
It is released under Apache-2.0 with an added attribution condition requiring a notice and repository link when used in products or services面向 users outside the organization. It is in early development, and updates may break existing data.
Summary drafted from the project's own website. Every sentence is backed by text on that page and was reviewed before publishing.