Your AI agent's memory, in an Obsidian vault

agentcairn stores your coding agent's memory as plain Markdown in an Obsidian vault you own — the source of truth, not a one-way export.

The memory is readable and editable Markdown with frontmatter and [[wikilinks]]. Open it in Obsidian, read what your agent knows, and correct it directly. Hand-edit a fact and the agent honors it on the next reindex — your edits are not overwritten.

Because the vault is the truth, you are never reverse-engineering a database to find out what your agent remembers. You just read the files.

The agentcairn-obsidian plugin surfaces your memory inside Obsidian: a filterable list of notes, with provenance (where each memory came from) and currency (how fresh it is). You see what the agent knows, where it learned it, and whether it is still current.

A rebuildable DuckDB index reconciles to the vault and provides fast hybrid retrieval (BM25 + vector + RRF, with an optional cross-encoder reranker). The index is a disposable cache; the vault is always the truth. cairn reindex rebuilds it from the Markdown with zero data loss.

This works with any host. Plugin hosts like Claude Code and Codex get a first-class plugin; every other MCP host is wired in with cairn install <host> (or cairn install --all for every detected agent). They all share one vault.

What is agent memory? · How agentcairn compares