Check Connector

Verify Nowledge Mem is running and guide the user to the best connector for their agent.

When to Use

• User asks about Nowledge Mem setup or configuration
• Memory tools are failing or not available
• User asks "is my memory working?" or "how do I set up Nowledge Mem?"
• First time using Nowledge Mem in this agent
• User asks about upgrading from skills to a native connector

Step 1: Check nmem CLI

nmem --json status

Use this rule before repairing anything:

• Same machine as Nowledge Mem Desktop: use the desktop app's bundled nmem. Ask the user to open Nowledge Mem; if the command is still missing, use Settings → Preferences → Developer Tools → Install bundled CLI.
• Different machine: install the standalone PyPI package nmem-cli on this machine, then point it at the user's Mem server. This covers remote servers, dev boxes, CI runners, hosted agents, and SSH machines. Python 3.11+ is required.

Do not install the PyPI CLI over the desktop-bundled CLI on the user's own desktop unless the user explicitly asks for a standalone CLI.

If the agent is not on the desktop machine and nmem is missing, install the PyPI CLI:

python3 -m pip install --user nmem-cli
# or: pipx install nmem-cli
# or for one-off checks: uvx --from nmem-cli nmem --json status

nmem config client set url https://<their-mem-server>
nmem config client set api-key nmem_...
nmem --json status

If nmem exists but status fails, Nowledge Mem is not reachable from this machine. Guide the user:

• Local desktop: open the Nowledge Mem app, then retry nmem --json status
• Remote/client machine: verify the URL/API key with nmem config client show
• Full install guide: https://mem.nowledge.co/docs/installation
• Remote access guide: https://mem.nowledge.co/docs/remote-access

Step 2: Detect Agent and Recommend The Best Path

For a fresh install, start from the universal connect skill:

Read https://mem.nowledge.co/SKILL.md and follow the instructions to install or update Nowledge Mem for the AI tool I am using.

Use this check-integration skill as the diagnostic fallback: confirm what was installed, explain what behavior the user should expect, and repair configuration when something failed.

Do not only answer "install X". Explain the behavior contract the user will get:

1. what starts automatically
2. what is only guided by skills/rules and still model-driven
3. whether threads are captured automatically, saved explicitly, or only supported as handoff summaries

Use this priority order:

1. Native connector first when the host has one
2. Universal Agent Plugin when the host supports Agent Plugins but has no native connector
3. Direct MCP when the host supports MCP but has no better package path
4. Reusable package when the host supports shared skills/prompts but has no native connector, Agent Plugins support, or MCP tool surface

Fresh users care about outcome, not transport. Tell them what they will actually get after setup.

Autonomy levels

Check which agent you're running in and recommend the native connector if available.

The canonical source for this table is community/integrations.json.

For Hermes updates, run the setup command from the guide and confirm it prints Thread import endpoint: /threads/import. If it prints an old endpoint or no endpoint, the user is likely launching Hermes with a different HERMES_HOME; set that explicitly and rerun setup before restart.

If the agent is not listed above:

• use the Universal Agent Plugin when the host supports Agent Plugins
• otherwise use direct MCP plus the recommended prompt block
• use the shared npx skills package when the host supports shared skills but has no MCP tool surface

Do not describe raw MCP as equivalent to a native connector.

Step 3: Verify

After setup, verify with:

nmem --json m search "test" -n 1

If this returns results (or an empty list with no error), the connector is working.

Then state the expected outcome in plain language:

• Working Memory: automatic, guided, or manual
• Recall: automatic, guided, or manual
• Distill: automatic, guided, or manual
• Threads: automatic capture, explicit save, handoff-only, or none

If the path is only guided, say what strengthens it:

• keep nmem available locally
• restart the host after install
• merge the package AGENTS.md or equivalent repo guidance when recommended
• configure the local nmem client in remote mode

What Native Connectors Add

Skills give you CLI-based memory access. Native connectors usually add:

• Auto-recall: relevant memories injected before each response (no manual search needed)
• Auto-capture: conversations saved as searchable threads at session end
• LLM distillation: key decisions and insights extracted automatically
• Graph tools: explore connections, evolution chains, and entity relationships
• Working Memory: daily briefing loaded or injected at session start
• Slash commands: /remember, /recall, /forget (where supported)

Do not promise all of these on every host. Match the actual path the user is setting up.

Links

• All Connectors
• Documentation
• Discord Community