Home › Getting Started
Getting Started
Work through the steps in order, and you will experience the basics end to end — from your first sign-in to saving a memory from your AI, recalling it, and confirming it on the dashboard.
Time required: about 10 minutes · What you need: a Google or GitHub account, and an MCP-capable AI service (Claude, ChatGPT, Cursor, Gemini, Codex, etc.)
1 Sign in
Open memory.kagura-ai.com and the sign-in screen appears.
- Choose "Continue with Google" or "Continue with GitHub".
- Check "I agree to the Terms of Service and Privacy Policy".
- Approve on the provider's authentication screen, and your workspace dashboard opens.
The Trial plan is invite-based. See the pricing page for current sign-up availability.
2 Connect your AI service
You use Kagura AI as a shared memory from the AI service you already work with. Open "API keys" or "OAuth apps" in the left menu to reach the "Credentials & Integrations" screen, where you find your connection credentials and the MCP connection URL (MCP endpoint) shared by all clients.
There are two ways to connect — pick the one that matches your AI service.
Option A: connect with an API key (programmatic access — Claude Code, Python SDK, etc.)
- On the "API keys" tab, issue a new API key. It is shown only once, so copy and save it.
- In your tool's MCP (connector) settings, set the MCP connection URL and the key you saved.
- Once connected, your AI can use the remember and recall tools.
Option B: connect with an OAuth app (desktop / web clients — Claude Desktop, ChatGPT)
- On the "OAuth apps" tab, create the OAuth app for your service (one for Claude, one for ChatGPT).
- Add the MCP connection URL to Claude Desktop's or ChatGPT's connector (MCP server) settings.
- When you connect, a Kagura AI sign-in screen opens; approve it, and your AI can use the remember and recall tools.
3 Save a memory from your AI service (remember)
In a conversation with your connected AI, simply ask it to remember what matters.
The AI calls remember and saves it, with a summary and tags, into a context — a "container" for memories. Keeping one context per project or theme makes things easier to find later (you can ask your AI: "Create a new context called 'Reading notes'").
"Remember this design decision, with the reasoning." / "Remember the conclusions from today's meeting."
4 Recall it from your AI service (recall)
Open a new conversation and try recalling what you just saved.
The AI searches the context with recall and pulls in only the relevant memories. No more re-explaining from scratch across conversations and tools.
"Recall how we decided this before." / "Review this with last week's discussion in mind."
5 Confirm it on the dashboard
Check in the web interface that your memories were saved. Go back to memory.kagura-ai.com and use the menu on the left.
- Dashboard — if total memories and API calls have grown, your AI service is connected and working. You can also see the memory-creation timeline.
- Contexts — shows each context's memory count and last activity. What you saved in step 3 appears here.
- Storage — check how much of your storage you are using.
6 Use it as a team
Everything you just did can be shared with your team as is. Invite members to your workspace and give each a role (owner / admin / member / viewer).
Each context has a visibility setting (private / shared / public), so shared team knowledge and personal memories stay cleanly separated. Save your team's decisions into a shared context, and every member gets the same knowledge — from whichever AI service they ask.