recalla-obsidian

search and chat your vault, without leaving obsidian

the obsidian plugin for recalla — a private, local-first, agent-native second brain. bm25 + semantic search and chat over your notes, all on your machine.

install

desktop only (the plugin manages a local engine process). the community directory listing is pending, so install manually from the latest release:

1. download main.js, manifest.json, and styles.css from the release.

2. create the folder <your vault>/.obsidian/plugins/recalla/ and put the three files inside it, keeping the filenames.

3. in Obsidian: Settings → Community plugins → turn off Restricted Mode.

4. reload Obsidian, then enable Recalla in the plugin list.

run recalla: make my vault agent-readable from the command palette. the plugin downloads the matching OS and CPU build from the pinned engine v0.5.0 release. it requires an embedded trusted digest, requires the published sidecar to name that exact asset, and hashes the bytes before execution. then it indexes the vault and starts the local api. no python or terminal setup is required.

managed builds support macOS arm64/x64, Linux arm64/x64, and Windows x64. advanced users on another architecture can set a custom recalla executable in plugin settings instead.

Related MCP server: Obsidian MCP Server

what it does

• search your vault with bm25 + semantic ranking, inside obsidian.

• chat over full matching sections with citations. without an llm key, recalla returns an offline extractive answer from your notes.

• copy a ready-to-paste mcp config so agents can read and write memory locally.

• manage the engine, indexing, and local server from the command palette.

• no cloud, no account, no telemetry. only context sent to a configured cloud llm leaves the machine.

the recalla ecosystem