One memory for every AI agent you use.

Any agent that speaks MCP or HTTP can save to it and recall from it. Start free.

  1. Monday 10:14 · coding agent in your terminal

    We chose Postgres for billing because we need transactions. Remember that.

    uteke_remember tags: decision, billing

    saved

  2. Thursday · chat assistant on your phone

    Why did we pick Postgres for billing?

    uteke_recall 1 result

    Billing uses Postgres: it needs transactions.

    written by
    coding agent
    saved
    Monday 10:14
    from
    work laptop
Two agents, two devices, one memory. Every memory says who wrote it, when and from where.

Agents change. Your memory should not.

You try a new assistant, switch models, add a coding agent. Memory built into one app stays in that app when you move on, so you explain yourself again. Uteke keeps what your agents learn in one place that belongs to you, and every agent you connect reads the same memory.

Three steps. No code.

  1. Add the URL. In your agent's connector or MCP server settings, add a server named Uteke with this URL: https://uteke.app/mcp No client ID and no secret. An agent that cannot sign in uses an API key from the portal instead.
  2. Sign in. When your agent asks, sign in with your CodeCora ID. Your first sign-in creates your Uteke account on Free.
  3. Work as usual. Say Remember that I prefer short answers. Later, in another session or another agent, ask How do I like my answers?

Works with any MCP client that connects to a remote server, and anything that can call HTTPS. Step-by-step guides for Claude, ChatGPT, Claude Code and Cursor.

Connect your agent

What people use it for.

A personal assistant that remembers
Preferences, people and recurring tasks, available in every chat and on every device.
Projects and decisions
Why a choice was made, kept across chats and across agents.
Coding agents that share lessons
Conventions and gotchas written once and read by every agent on the project.
Research notes and documents
Notes and documents you can recall by meaning, not only by keyword.
Your own agents
Builders use the REST API and MCP to give their agents a memory.

How it differs from what you use now.

Memory built into an AI app
Lives inside one vendor's app. When you switch agents, it stays behind.
A notes app
Made for people to read. An agent has to load whole files to find one fact.
A memory API that runs a model
Pays for model calls to store and find things, and keeps a model's summary instead of what was said.
Uteke Cloud
One memory for every agent, found by search, not a model. You can see, edit and export every memory.

What we promise.

Your data, your exit
Export everything as NDJSON whenever you want, and run the same engine on your own machine with open-source Uteke.
Secrets masked
API keys, tokens and private keys are masked before a memory is stored.
No AI on recall
No AI reads your memories when they are recalled. Recall is search, not a model. Optional features that use a model are off by default, name the provider that runs them, and never run on recall.
See and edit every memory
Your agent or the API can list, read, correct and delete any memory you stored.
Change history
Edits and replacements are recorded with who made them, and a replaced memory stays linked to the one that replaced it.
One region
Your memories are stored and searched in one region, Singapore, on servers we rent and operate. See the privacy policy.

Open source you can check.

The engine behind Uteke Cloud is open-source Uteke. Read how recall works, run it on your own machine, and compare the results with what the cloud gives your agents.

Free to start, in open beta.

Sign in with a CodeCora ID and your account is ready on Free, no card: 500 memories, 150 recalls a day, and export whenever you want. Paid plans come later.

Plans and limits

For developers: one URL, one key.

The same memory is open to your own tools: any MCP client through 45 tools, or plain JSON over HTTPS at https://uteke.app/api/v2. A command-line agent connects in one line:

claude code
claude mcp add --transport http uteke https://uteke.app/mcp \
  --header "Authorization: Bearer utk_live_…"

remember

POST /remember
{ "content":
  "Payments need idempotency keys",
  "tags": ["backend"] }

Written once. Embedded and indexed on arrival.

recall

POST /recall
{ "query":
  "how do we avoid double charges?" }

Keyword and vector results fuse into one ordering. No mode to pick.

export

GET /export
manifest
memories
documents

Every memory and document with its tags, as one NDJSON stream. Leave whenever you want: no export fee, no lock-in.

Rather run it yourself? Open-source Uteke is the same engine with the same tool names, on your own machine.

Developer guide

Questions.

Which agents work with Uteke?

Any MCP client that can connect to a remote server, and any program that can call HTTPS. The connection guide has exact steps for Claude, ChatGPT, Claude Code and Cursor.

Do I need to write code?

No. Add one URL to your agent and sign in. The REST API is there if you want to build on it.

Does an AI read my memories?

No AI reads your memories when they are recalled: recall is search. Optional features that use a model are off by default, name the provider that runs them, and never run on recall.

What if I want to leave?

Export everything as NDJSON from the portal or the API (4 exports a month on Free), then delete your account yourself. The same engine runs on your own machine as open-source Uteke.

Where is my data stored?

In one region, Singapore, on servers we rent and operate. The privacy policy lists every provider that may process your data.

What happens to passwords or keys I paste?

Text that looks like an API key, token or private key is masked before the memory is stored.

What does it cost?

Uteke Cloud is free in open beta, with no card. Paid plans come later; what Free includes and what happens at a limit are on the pricing page.

Not sure it fits? Ask your AI.

Paste this into the assistant you already use. It reads our llms.txt, a plain summary written for AI agents, and answers for your situation.

Read https://uteke.app/llms.txt. Then, based on how I use AI agents, tell me in plain words whether Uteke Cloud would help me, which agent I should connect first, and what it would not do for me.

Every agent, one memory you can take with you.

Uteke Cloud is in open beta. Create an account with a CodeCora ID, the sign-in for CodeCora apps, and start on Free. Uteke receives only your email address and an account ID. Questions go to the address below.