nowledge-mem-copilot-cli-plugin

Nowledge Mem — Copilot CLI Plugin

Your personal knowledge graph, built into GitHub Copilot CLI. Copilot loads your Working Memory, nudges for recall, and captures sessions automatically.

Install

# Add the Nowledge community marketplace
copilot plugin marketplace add nowledge-co/community

# Install the plugin
copilot plugin install nowledge-mem@nowledge-community

Prerequisite: nmem CLI must be in your PATH:

pip install nmem-cli    # or: pipx install nmem-cli
# Arch Linux: yay -S nmem-cli  # or: paru -S nmem-cli
nmem status             # verify connection

On Windows/Linux with the Nowledge Mem desktop app, nmem is already bundled.

Using Copilot CLI inside WSL? Paste this into your WSL terminal to bridge nmem:

mkdir -p ~/.local/bin && cat > ~/.local/bin/nmem << 'SHIMEOF'
#!/bin/bash
python3 - "$@" <<'PY'
import subprocess
import sys

cmd = subprocess.list2cmdline(["nmem.cmd", *sys.argv[1:]])
raise SystemExit(subprocess.run(["cmd.exe", "/s", "/c", cmd]).returncode)
PY
SHIMEOF
chmod +x ~/.local/bin/nmem

This calls the Windows nmem via interop — no extra setup or network configuration needed.

What You Get

Automatic (no action needed):

• Working Memory briefing loaded at every session start, resume, and clear
• Per-turn behavioral nudge with memory search and save syntax
• Session conversations captured to your knowledge graph on each response
• Current transcript captured before compaction, then Working Memory recovered after compaction events

Skills (model-mediated; Copilot may expose them as namespaced entries):

• Read Working Memory — loads Context Bundle when available, or your daily context briefing as the lightweight fallback
• Search Memory — searches both distilled memories and prior sessions when continuity matters
• Distill Memory — suggests saving breakthroughs and decisions
• Save Thread — saves a concise summary thread when the user explicitly asks to save/checkpoint the session

The plugin no longer ships separate command docs. Skills are still interpreted by the model and then invoke nmem as needed; they are not host-native direct shell commands. For direct terminal checks outside the skill surface, run the nmem CLI yourself (for example nmem status).

How It Works

Lifecycle Hooks

Event	Trigger	Action
SessionStart	New, resume, or clear	Loads Context Bundle via nmem --json context --source-app copilot-cli, then falls back to Working Memory
SessionStart	After compaction	Re-loads Context Bundle or Working Memory + checkpoint prompt
UserPromptSubmit	Every user message	Injects search/save syntax as context
Stop	Model finishes responding	Captures session to knowledge graph (async)
PreCompact	Before context compaction	Captures the current transcript before context is compressed
SessionEnd	Session exits or clears	Runs the same idempotent capture path as a final backstop

The SessionStart hook tries nmem --json context --source-app copilot-cli first so Copilot receives identity, active scope, active rules, and current priorities when the installed CLI supports it. It falls back to nmem --json wm read, then to ~/ai-now/memory.md only as the Default-space compatibility path.

The Stop, PreCompact, and SessionEnd hooks run the same Python capture script from the plugin's own hooks/ directory. For older installs, it still falls back to ~/.copilot/nowledge-mem-hooks/ if that compatibility copy exists. It reads the Copilot CLI transcript, extracts messages, filters secrets, and creates threads via nmem t import. This is idempotent — repeated runs only append new content.

Session Capture Details

The capture script (copilot-stop-save.py) handles:

• Secret filtering — redacts API keys, tokens, credentials (6 patterns + 10 skip patterns)
• Incomplete turn detection — skips when waiting for user input or background tasks
• Concurrent safety — file locking prevents race conditions with parallel sessions
• Auto-distill — valuable sessions are automatically distilled with guardrails (cooldown, content hash dedup, minimum thresholds)
• Thread ID — copilot-{session_id} (stable per-session, enables incremental append)

Local vs Remote

The plugin works transparently in both modes:

• Local (Mem on same machine): Working Memory read from Mem, sessions captured locally.
• Remote (Mem on different machine): configure this machine once with:

nmem config client set url https://your-server
nmem config client set api-key your-key

That writes the shared local client config used by nmem and the plugin. You can also use environment variables (NMEM_API_URL, NMEM_API_KEY) for temporary overrides.

Spaces

Spaces are optional. If one Copilot CLI process naturally belongs to one project or agent lane, launch Copilot CLI with:

NMEM_SPACE="Research Agent"

The session-start Context Bundle / Working Memory read, per-turn guidance, skills, and background capture will then stay in that lane automatically.

For multi-agent orchestrators, set NMEM_AGENT_ID=<agent-slug> per spawned Copilot CLI worker. Add NMEM_SPACE only when that run should override the AI Identity's default space. NMEM_HOST_AGENT_ID is for advanced host-id aliases. Context Bundle will use the stable identity while keeping source_app=copilot-cli for provenance.

Update

copilot plugin marketplace add nowledge-co/community
copilot plugin marketplace update nowledge-community
copilot plugin update nowledge-mem

Restart Copilot CLI to apply changes.

Customize without editing the plugin

Copilot CLI already has a native instruction layer. Use that as your override path.

• Shared repo rules: .github/copilot-instructions.md or .github/instructions/*.instructions.md
• Personal rules across repos: ~/.copilot/copilot-instructions.md
• Do not edit installed plugin hooks or scripts under ~/.copilot/installed-plugins/...

Keep the plugin for lifecycle hooks and capture. Put your custom memory behavior in Copilot instruction files so updates do not wipe it out.

Troubleshooting

nmem not found: Install with pip install nmem-cli or pipx install nmem-cli. If you're in WSL, see the WSL setup above.

Server not running: Start the Nowledge Mem desktop app, or run nmem serve on your server

Session capture not working: Restart Copilot CLI first. If capture still does not run, check ~/.copilot/nowledge-mem-hooks/hook-log.jsonl for diagnostics. For older installs or local-development setups, you can still run scripts/install-hooks.sh from the source checkout as a compatibility fallback.

Check status: Run nmem status to see connection details

Links

• Documentation
• Discord
• GitHub

---

Made with care by Nowledge Labs