Mem9 setup

Skill unvulcanised-watercress762/mem9/claude-plugin/skills/mem9-setup

Enable AI agents to retain memory across sessions using persistent storage designed for continuous context retention.

Install

npx -y skills add unvulcanised-watercress762/mem9 --skill mem9-setup

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Setup Mem9 persistent memory. Triggers: set up mem9, install mem9, configure memory, enable memory.

SKILL.md

2.9 KB, as published. Nobody here has run it

Mem9 Setup for Claude Code

You are an automated installer. Execute ALL steps yourself — do NOT ask the user to run anything except the final restart.

Step 1: Check if already configured

python3 -c "
import json, os, sys
path = os.path.expanduser('~/.claude/settings.json')
if not os.path.exists(path):
    print('NO_SETTINGS')
    sys.exit(0)
with open(path) as f:
    cfg = json.load(f)
tid = cfg.get('env', {}).get('MEM9_TENANT_ID', '')
if tid:
    print(f'ALREADY_CONFIGURED:{tid}')
else:
    print('NEEDS_TENANT')
"

If ALREADY_CONFIGURED:<id> → Skip to Step 3.
If NEEDS_TENANT or NO_SETTINGS → Continue to Step 2.

Step 2: Provision tenant and write config

2a: Provision a new tenant

curl -s -X POST https://api.mem9.ai/v1alpha1/mem9s

Extract the id field from the JSON response. This is the MEM9_TENANT_ID.

If the curl fails, tell the user the API might be down and ask them to try later.

2b: Write tenant ID to settings.json

Read ~/.claude/settings.json (create if missing), merge MEM9_TENANT_ID into the env object, and write it back. Preserve all existing settings.

python3 -c "
import json, os, sys

tenant_id = sys.argv[1]
path = os.path.expanduser('~/.claude/settings.json')

# Read existing or start fresh
cfg = {}
if os.path.exists(path):
    with open(path) as f:
        cfg = json.load(f)

# Merge env
env = cfg.get('env', {})
env['MEM9_TENANT_ID'] = tenant_id
cfg['env'] = env

# Write back
os.makedirs(os.path.dirname(path), exist_ok=True)
with open(path, 'w') as f:
    json.dump(cfg, f, indent=2)
    f.write('\n')

print(f'OK: MEM9_TENANT_ID={tenant_id} written to {path}')
" "REPLACE_WITH_ACTUAL_TENANT_ID"

Replace REPLACE_WITH_ACTUAL_TENANT_ID with the actual tenant ID from Step 2a.

Step 3: Install plugin via CLI

These are CLI commands — run them directly in Bash:

claude plugin marketplace add mem9-ai/mem9

claude plugin install mem9@mem9

If claude plugin marketplace add fails with "already exists", that's fine — skip it and continue to install.

If claude plugin install fails with "already installed", that's fine — the plugin is already set up.

Step 4: Confirm and tell user to restart

Summarize what was done:

✅ Tenant provisioned: <tenant_id>
✅ ~/.claude/settings.json updated with MEM9_TENANT_ID
✅ Plugin marketplace added and plugin installed

Then tell the user:

All done! Just restart Claude Code to activate Mem9 memory.

After restart, verify by saying "remember that this project uses React 18", then start a new session and ask "what UI framework does this project use?" — the agent should recall it.