Local MCP server and Kimi Work skill for turning a Familiar Markdown vault into a second brain.
Familiar Second Brain connects three local pieces:
- A Markdown vault at
~/Documents/kimi/workspace/familiar-vault. - A Kimi Work skill that saves session summaries and transcripts into that vault.
- A stdio MCP server that lets MCP clients save, search, read, and write vault notes.
The result is simple: when you ask Kimi Work or another MCP-capable app to use familiar, it can work against the same local Markdown brain instead of scattering context across desktop files.
| Capability | Where it lives | Output |
|---|---|---|
| Save durable memories | save_memory MCP tool and Kimi skill |
Markdown notes in _Inbox/ or project folders |
| Search the brain | search_memory MCP tool |
Ranked Markdown matches with matched fields and context windows |
| Read and write notes | read_note, write_note MCP tools |
Plain .md files inside the vault |
| Capture Kimi sessions | summarize_sessions.py |
Daily/Kimi Sessions/ and Daily/Kimi Transcripts/ |
| Generate daily brain briefs | brain_brief.py |
Daily/YYYY-MM-DD Brain Brief.md |
| Triage inbox notes | inbox_triage.py |
suggested moves, tags, links, and optional safe moves |
| Generate project briefs | project_briefs.py |
Projects/<project>/Brief.md |
| Audit vault quality | vault_audit.py |
duplicate and contradiction reports |
| Resurface useful notes | resurface.py |
Daily/Resurfaced/YYYY-MM-DD Resurfaced Notes.md |
| Share with apps | MCP configs | Kimi Work, Codex, Claude, Cursor |
Run the installer in dry-run mode first:
cd /Users/username/Downloads/familiar-second-brain
/usr/bin/python3 scripts/install.py --dry-runInstall into the current local Kimi/Familiar setup:
/usr/bin/python3 scripts/install.pySmoke-test the MCP server:
make smokeRun the full test suite:
make testfamiliar-second-brain/
src/
familiar_mcp_server.py # stdio MCP server
kimi_skill/
SKILL.md # Kimi Work behavior guide
scripts/
save_to_familiar.py # quick note capture helper
summarize_sessions.py # Kimi session summary and transcript capture
brain_brief.py # daily brief from sessions, transcripts, inbox notes
inbox_triage.py # suggestion-first inbox organization
project_briefs.py # per-project memory brief generation
vault_audit.py # read-only duplicate and contradiction audit
resurface.py # recurring useful-note resurfacing
tests/
scripts/
install.py # local installer and config writer
smoke_mcp.py # stdio MCP smoke test
docs/
architecture.md
installation.md
operations.md
security.md
assets/*.svg # animated diagrams
The server exposes these tools:
save_memory: create a Markdown note in the vault.search_memory: search Markdown notes by keyword with title, tag, heading, link, and body scoring.read_note: read a note by relative path.write_note: write a Markdown note by relative path.list_recent_notes: list recently modified notes.capture_kimi_sessions: run the Kimi session capture job.vault_status: report vault paths and note count.
Default install paths:
Vault:
~/Documents/kimi/workspace/familiar-vault
MCP server:
~/Documents/kimi/workspace/familiar-vault/.familiar/mcp/familiar_mcp_server.py
Kimi skill:
~/Library/Application Support/kimi-desktop/daimon-share/daimon/skills/familiar-second-brain
Kimi generated MCP config:
~/Library/Application Support/kimi-desktop/daimon-share/daimon/runtime/kimi-code/home/mcp.json
The repo deliberately does not include personal vault notes, Kimi transcripts, API keys, or app databases.
This project uses only the Python standard library.
make test
make smokeFor MCP protocol debugging, run:
/usr/bin/python3 scripts/smoke_mcp.py \
--server src/familiar_mcp_server.py \
--vault ~/Documents/kimi/workspace/familiar-vault