Junê

Junê gives an AI assistant a shared, cited knowledge-graph memory: ask questions, search, remember facts and keep standing instructions across sessions.

What you can do

• Ask cited questions against the bound canvas, optionally using a bring-your-own LLM key
• Search and read the knowledge graph and pages in the canvas
• Remember new information and maintain standing docs, skills and learnings
• Re-inject a periodic digest of standing docs into tool results so context persists
• Optionally export and import pages to a local directory for repo sync, and upload files for ingestion when those roots are configured

When to use it

Useful when an assistant should carry durable, citable memory between sessions or share a knowledge base with a team, rather than starting from zero each conversation. Also fits setups that want standing instructions or skills kept in one place and refreshed automatically.

Requirements

Runs locally over stdio, launched with uvx from the june-mcp PyPI package, so Python and uv are needed. Requires JUNE_BASE_URL pointing at a June endpoint (the desktop app's local engine, a self-hosted service, or a hosted Team workspace) and JUNE_CANVAS naming the canvas to bind to. JUNE_API_KEY is required unless JUNE_ALLOW_ANON=1 is set; JUNE_LLM_KEY is optional. Desktop only.