FAQ & troubleshooting

About Kijito

Does Kijito train or modify the underlying model?

No. Kijito is an external memory layer. It stores and organizes experience your agent records and recalls; it does not change the model's weights or fine-tune it. The model keeps whatever it learned in training and adds durable, revisable memory on top.

What does Kijito mean by "experience"?

Retained outcomes that can affect later work: decisions and their reasons, what a result taught, corrections to earlier beliefs, and unfinished work that survives a session boundary. Kijito makes that experience persistent and retrievable - it does not claim to generate experience on its own.

What is automatic dreaming versus deep dreaming?

Automatic (scheduled) consolidation runs on a schedule and does structural, content-blind maintenance: it decays weak material and keeps the graph organized without reading your content. When an authorized content key is available and the configuration permits, it can also organize by meaning. Deep dreaming is agent-invoked: Kijito hands an agent a prioritized review worklist, and the agent verifies, connects, corrects, decomposes, or fades what it finds using explicit memory tools. See Concepts.

Which OpenAI products are currently verified?

None are verified yet. Today the verified paths are Claude Code and the REST API. Claude's desktop/web MCP connector and Codex CLI are in testing; the ChatGPT connector and OpenAI Apps are planned. The Connections page carries the current per-client status.

Can a local or open-weight model use hosted Kijito?

Yes, through any MCP-capable host or the REST API. But running a local model does not make the hosted service operator-blind - the memory service is still hosted, so it stays encrypted at rest, not operator-blind. The operator-blind path is running the decryption off our servers - via Enterprise on-premises, or a local helper you run - both on our roadmap.

What happens if Kijito can't reach my agent?

Nothing is lost. Hive messages are durable and stay in the mailbox after they are sent. Urgent delivery sends an opaque wake signal, but the recipient still pulls the message from Kijito - so if your agent is offline, it retrieves the message when it comes back. Delivery of the wake and consumption of the message are separate; "sent" does not mean an agent was resumed.

Why are there so many MCP tools?

Kijito exposes a full toolset over MCP because memory has many verbs - remember, recall, correct, update, fade, relate, dream, hive send/inbox, presence, and more. Day to day an agent reaches for a small core; the rest are there when a workflow needs them.

Connecting

My MCP client won't connect

Work down this list:

• Use the exact Streamable HTTP URL from the Quickstart connector (it ends in /mcp). A trailing-slash or /sse variant will not work - Kijito serves a single modern MCP transport.
• Complete the OAuth sign-in when the client opens a browser. If nothing opens, check that pop-ups aren't blocked, then retry the connect.
• In Claude Code, confirm both files exist in your project root - .mcp.json with the server block and .claude/settings.local.json with {"enableAllProjectMcpServers": true} - then restart the session. Missing the second file is the most common reason the tools never appear.
• Restart the client after adding the server so it re-reads its connector config.

I connected, but the memory tools don't show up

Almost always a config-not-reloaded issue: restart the client/session so it picks up the new MCP server, and make sure you finished the sign-in step (an unauthenticated connection advertises no tools). If you're on the REST API instead, the "tools" are just the endpoints in the API reference.

Authentication

My API calls suddenly return 401

Access tokens are JWTs valid for one hour. A 401 after a while usually means the token expired. Mint a fresh one with your refresh token at POST /api/v1/auth/refresh (or sign in again) and retry. For long-running or unattended agents, prefer the MCP connector, which handles its own OAuth sign-in and refresh - you don't manage tokens by hand. Or, for direct REST, mint a standalone API key - a scoped kjt_ bearer token that never expires unless you set it to - via POST /api/v1/auth/tokens (see the API reference); it sidesteps the hourly refresh entirely.

Do I need to log in for every request?

No. Log in once to get the token, then send it as an Authorization: Bearer header on each call until it expires, and refresh when it does. MCP clients keep their own session, so you sign in once per client.

Recall behaviour

Why didn't it recall something I know I saved?

A few things to check, roughly in order of likelihood:

• Recall is by meaning, not keywords. If your query is worded very differently from the memory, similarity may be low. Ask in terms closer to the idea, or raise the limit. Writing memories in the words a future search would use makes them far more findable.
• It may have faded. Low-importance, rarely-touched memories decay on the forgetting curve and drop out of recall over time. Give important memories a higher importance, or mark them permanent, so they persist.
• Scope or project mismatch. If the memory was stored under a specific project or a narrower scope than you're recalling with, it may be filtered out. Recall at the account (global) scope to be sure.
• It was corrected or faded. A corrected memory is superseded by its newer version; a faded one is intentionally demoted. That is the system working, not a bug.
• Different account. Memories never cross accounts. If you saved it while signed in as a different account, it won't appear.

See Concepts for how recall by meaning and the forgetting curve work.

It recalled an old fact that's no longer true

Ask your agent to correct it. Correction retires the stale version (keeping it for history) and writes the current one, so future recall surfaces the truth while the audit trail stays intact. Don't just delete - correcting preserves the record of how the understanding changed.

Using multiple clients

Can I connect Kijito to more than one app at once?

Yes - that's the point. One account is one memory graph, and you can connect Claude Code, Cursor, other MCP clients, and the REST API to it simultaneously. What one client remembers, another recalls. Point every client at the same server URL and sign in with the same account.

How do I keep different projects or agents separate?

Use personas and projects (see Concepts). A persona attributes work to a named identity, and a project sub-groups memories within it, so several agents can share one graph while it stays clear who wrote what and recall can be narrowed when you want it. Your agent sets the persona and project when it stores a memory, so no extra client configuration is required.

Account & limits

Why did I hit a memory limit so quickly?

New accounts start with a reduced cap until you verify your email. Verify to lift it to the full free-tier allowance. If you're seeing a 429/limit response and your email is already verified, you may have reached your plan's memory-node cap or the per-minute write-rate limit - check GET /api/v1/usage for where you stand, and fade or export memories you no longer need.

I didn't get the verification email

Check spam, and confirm the address on your account page. You can re-send the verification from there. Until it's verified the account works - it just keeps the lower memory cap.

Is my data really private?

Your content is encrypted at rest under a key derived from your password, and no account can read another's data. Read Security & trust for the exact model, including where the hosted service's trust boundary sits and how the operator-blind paths keep decryption off our servers.

Can I get my data out?

Any time. GET /api/v1/export returns your whole graph as JSON, and /api/v1/export/gexf gives you a Gephi-ready file. Keeping your own exports of anything important is a good habit - Kijito is in alpha and provided as-is.

Still stuck?

Email [email protected] and we'll help. Found a bug in Kijito itself? Your agent can file it directly with the feedback tool.