Self-hosted MCP server + dashboard giving Claude Code, Cursor & Hermes agents shared long-term memory over a git-backed markdown vault. Agent-native, Obsidian-compatible second brain — markdown RAG, no vector DB.
TypeScript agent-memoryai-agentsanthropicclaude-codecursorgitknowledge-baseknowledge-graphllmmarkdownmcpmcp-servermodel-context-protocolnote-takingobsidianpkmragsecond-brainself-hostedzettelkasten
/**
* Anthropic models offered across Engram — the single source of truth shared by the
* Curator chat picker (client), the Settings capture-model dropdown (client), and the
* server (curator stream + brain_capture harness). Plain constants, no other imports,
* so it's safe in both client and server bundles and can't create an import cycle.
*/
export const CURATOR_MODELS = [
{ id: "claude-opus-4-8", label: "Opus 4.8" },
{ id: "claude-sonnet-5", label: "Sonnet 5" },
{ id: "claude-haiku-4-5", label: "Haiku 4.5" },
] as const;
/** Default model for the Curator chat (the strongest — you're reasoning over your brain). */
export const DEFAULT_CURATOR_MODEL = "claude-opus-4-8";
/** Default model for brain_capture auto-filing (cheap + fast — filing a rough note is a simple task). */
export const DEFAULT_CAPTURE_MODEL = "claude-haiku-4-5";
/** All supported model IDs, for validation. */
export const SUPPORTED_MODEL_IDS = new Set<string>(CURATOR_MODELS.map((m) => m.id));