Rust library for reading and writing Bear notes directly via the local SQLite database.
No network, no CloudKit, no Bear.app process required. Works while Bear is running or closed.
- macOS
- Bear installed (the database lives in Bear's app group container)
[dependencies]
bear-rs = "0.2"To also get the bear CLI and MCP server:
[dependencies]
bear-rs = { version = "0.2", features = ["cli"] }Or install the binary directly:
cargo install bear-rs --features cliuse bear_rs::{SqliteStore, store::ListInput};
// Read
let store = SqliteStore::open_ro()?;
let notes = store.list_notes(&ListInput::default())?;
let note = store.get_note(Some("7E635AD3-..."), None, true, true)?;
// Write
let store = SqliteStore::open_rw()?;
let note = store.create_note("# My Note\n\nHello world", &["work", "rust"], false)?;
store.append_to_note(Some(¬e.id), None, "more text", Default::default(), true, Default::default())?;
store.trash_note(Some(¬e.id), None)?;After any write, bear-rs posts the Darwin notification net.shinyfrog.bear.cli.didRequestRefresh so Bear's UI refreshes automatically.
Open with SqliteStore::open_ro() for reads or SqliteStore::open_rw() for writes.
Reading
| Method | Description |
|---|---|
list_notes(input) |
List notes with optional tag filter, sort, and limit |
get_note(id, title, attachments, pins) |
Fetch a single note by ID or title |
cat_note(id, title, offset, limit) |
Raw note text with optional pagination |
search_notes(query, limit) |
Bear search syntax (@todo, #tag, -negation, etc.) |
search_in_note(id, title, string, ignore_case) |
Line matches within a single note |
list_tags(id, title) |
All tags, or tags for one note |
list_pins(id, title) |
Pin contexts for one or all notes |
list_attachments(id, title) |
Attachments for a note |
read_attachment(id, title, filename) |
Raw attachment bytes |
Writing
| Method | Description |
|---|---|
create_note(text, tags, if_not_exists) |
Create a note; title extracted from first line |
append_to_note(id, title, content, position, update_modified, tag_position) |
Append or prepend text |
write_note(id, title, content, base_hash) |
Overwrite note content; optional hash guard |
edit_note(id, title, ops) |
Find/replace operations |
trash_note(id, title) |
Move to trash |
archive_note(id, title) |
Archive |
restore_note(id, title) |
Restore from trash or archive |
add_tags(id, title, tags) |
Add tags |
remove_tags(id, title, tags) |
Remove tags |
rename_tag(old, new, force) |
Rename tag across all notes |
delete_tag(name) |
Delete tag and remove from all notes |
add_pins(id, title, contexts) |
Pin in contexts ("global" or tag name) |
remove_pins(id, title, contexts) |
Unpin |
add_attachment(id, title, filename, data) |
Attach a file |
delete_attachment(id, title, filename) |
Mark attachment unused |
search_notes accepts Bear's query syntax:
@today modified today
@yesterday modified yesterday
@lastNdays modified in last N days
@date(YYYY-MM-DD) modified on a specific date
@todo has incomplete todos
@done has completed todos
@pinned pinned notes
@tagged has at least one tag
@untagged has no tags
@images has images
@files has files
@code has code blocks
@locked locked notes
#tag has tag
-word does not contain word
"exact phrase" exact phrase match
use bear_rs::export::{ExportNote, export_notes};
let notes: Vec<ExportNote> = store.list_notes(&Default::default())?
.into_iter()
.map(Into::into)
.collect();
export_notes("./output".as_ref(), ¬es, true, true)?;Enable the cli feature to build the bear binary and access the MCP server:
cargo install bear-rs --features clibear list
bear list --tag work --sort modified:desc --format json
bear show --title "My Note" --fields all,content
bear search "@today @todo"
bear create "My Note" --content "Body" --tags work
bear append --title "My Note" --content "More text"
bear write <id> --content "# Title\n\nBody"
bear edit <id> --at "old" --replace "new"
bear trash --title "My Note"
bear archive <id>
bear restore <id>
bear tags list
bear tags add <id> work
bear tags rename old new
bear pin add <id> global
bear attachments list <id>
bear mcp-server
The MCP server speaks JSON-RPC 2.0 over stdin/stdout and exposes one tool per command. Configure it in Claude Desktop or any MCP-compatible client by pointing it at bear mcp-server.
cargo build
cargo test
cargo clippy --all-targets -- -D warnings # library only
cargo clippy --features cli --all-targets -- -D warnings # with CLI