From kagura-memory
Guides connection setup to Kagura Memory Cloud via MCP, checks status, and shows workflows for using memory contexts in solo or team projects.
How this command is triggered — by the user, by Claude, or both
Slash command
/kagura-memory:guideThe summary Claude sees in its command listing — used to decide when to auto-load this command
Show the user how to use Kagura Memory Cloud. If not connected, help them set up the MCP connection. ## Steps ### 1. Check MCP connection If this succeeds, MCP is connected. Skip to Step 3. If this fails, guide the user through setup (Step 2). ### 2. MCP connection setup (if not connected) Check for existing config: If no config exists, ask the user for: - **Server URL**: Where their Kagura Memory Cloud instance is running (default: `http://localhost:8080`) - **Workspace ID**: Found in the web UI URL bar after login - **API key**: Created at Workspace > Integrations > API Keys (s...
Show the user how to use Kagura Memory Cloud. If not connected, help them set up the MCP connection.
list_contexts()
If this succeeds, MCP is connected. Skip to Step 3.
If this fails, guide the user through setup (Step 2).
Check for existing config:
cat .mcp.json 2>/dev/null || echo "No .mcp.json found"
If no config exists, ask the user for:
http://localhost:8080)kagura_)Create or update .mcp.json:
{
"mcpServers": {
"kagura-memory": {
"type": "http",
"url": "{server_url}/mcp/w/{workspace_id}",
"headers": {
"Authorization": "Bearer {api_key}"
}
}
}
}
If .mcp.json already exists with other servers, merge the kagura-memory entry.
Add to .gitignore (contains API key):
grep -q '.mcp.json' .gitignore 2>/dev/null || echo '.mcp.json' >> .gitignore
Tell the user to restart Claude Code to pick up the config.
Solo developer — project knowledge base:
remember decisions, architecture choices, bug fixes as you workrecall at the start of each session to load contextTeam — shared knowledge base:
is_private: false)recall to learn project historyExternal integration — file and vault sync:
Use source_uri and source_type with remember to track where knowledge originated (e.g. Obsidian vaults, local files, web pages). Then use source_uri_prefix and source_type filters with recall to query memories from a specific source. Example: remember(..., source_uri="vault://my-vault/note.md", source_type="vault") → recall(..., filters={"source_uri_prefix": "vault://my-vault/"}).
Session workflow:
/kagura-memory:session-start to restore context/kagura-memory:remember and /kagura-memory:recall as needed/kagura-memory:session-summary to save learningsTo automatically remind yourself to restore session context, add this hook to your project's .claude/settings.json. Substitute {server_url} with the same URL you put in .mcp.json (e.g. https://memory.kagura-ai.com for the hosted service or http://localhost:8080 for a local instance):
{
"hooks": {
"SessionStart": [
{
"matcher": "",
"hooks": [
{
"type": "command",
"command": "curl -sf {server_url}/health >/dev/null 2>&1 && echo 'Kagura Memory Cloud is connected. Run /kagura-memory:session-start to restore previous session context.' || true"
}
]
}
]
}
}
Merge this into your existing hooks object if you already have other hooks defined.
| Skill | Description |
|---|---|
session-start | Restore previous session context |
session-summary | Save session knowledge before ending |
recall | Search past knowledge |
remember | Save new knowledge |
guide | This guide |
smoke-test | Verify all MCP tools work |
If you're setting up Kagura Memory Cloud plugin in a new project or on another machine:
Option A: Marketplace install (recommended)
Inside Claude Code, register the marketplace first (skip this line if it's already registered — /plugin marketplace list shows what's there), then install:
/plugin marketplace add kagura-ai/memory-cloud
/plugin install kagura-memory@kagura-memory-cloud
The install argument is <plugin-name>@<marketplace-name> — the plugin is kagura-memory and it lives in the kagura-memory-cloud marketplace.
After install, the plugin skills (/kagura-memory:*) become available. Proceed to Step 2 above to configure the MCP server connection.
Option B: Local install (from this repository)
If you have the memory-cloud repo cloned locally, first add it as a local marketplace, then install:
/plugin marketplace add /path/to/memory-cloud
/plugin install kagura-memory@kagura-memory-cloud
This installs the plugin from .claude-plugin/plugin.json and claude-skills/ in the repo.
After either install:
.mcp.json with workspace-specific credentialsnpx claudepluginhub kagura-ai/memory-cloud --plugin kagura-memory/guideDisplays quick start guide for vibe-guide plugin, covering session commands, workflows, examples, troubleshooting, and created files.
/guideRuns interactive first-time user guide: diagnoses project setup (detects files like package.json/go.mod, CLAUDE.md, Git), provides situation-specific advice, suggests workflows, marks onboarding complete.
/guideScans project docs, reviews, and exit status for testany-eng workflow; outputs current stage, ready baselines, Mermaid DAG, and 1-3 next skill or downstream recommendations.
/guideLoads and executes Hypo-Workflow skill instructions for guided planning. Supports user-provided arguments and enforces disciplined question-asking for scope, safety, and architecture decisions.