Bastra Recall
bastra-recall is a persistent memory layer that works across every AI surface — Claude (Code, Desktop, Web via Custom Connector), ChatGPT (via Custom GPT Actions over HTTP), Cursor, and anything else that speaks MCP or HTTP.
One local daemon serves all your AI tools at once. The vault is a folder of plain markdown files with YAML frontmatter — Obsidian-compatible, editable by hand, sync-friendly (iCloud / Google Drive / Dropbox).
Saves autonomously when a lesson is learned. Recalls before acting via PreToolUse and SessionStart hooks (Claude Code). Hybrid BM25 + vector search with the recall_when field as highest-weighted trigger — much better recall semantics than embed-everything approaches.
Memory types are typed (lesson, decision, preference, project-fact, workflow) with type-specific expiry defaults. Scope-based isolation per project. Wikilinks in bodies auto-mirror to a related[] graph.
MIT licensed, BYOK. Pre-alpha.
Server Config
{
"mcpServers": {
"bastra-recall": {
"command": "node",
"args": [
"/abs/path/to/bastra-recall/packages/daemon/dist/mcp-forwarder.js"
],
"env": {
"BASTRA_VAULT_PATH": "/abs/path/to/your/vault/memorys"
}
}
}
}