Agent-readable docs index: /docs/llms.txt. Full docs in one file: /docs/llms-full.txt. Download /docs/docs.zip to grep all markdown files locally.

Quickstart

Five minutes: get a key, connect something, store a fact, change it, and read back the current one with its source.
Install from npm (kaibo), PyPI (kaibo) or the CLI (curl -fsSL $KAIBO_URL/install.sh | sh). An AI agent can also connect over MCP, and anything can use the HTTP API with curl, with nothing to install.

1. Get a key

Open the dashboard (the /app page of the site you are reading this on), create a project and an API key. A key looks like kaibo_<id>.<secret> and is shown once. Choose the scope read and write for this tutorial; forgetting later needs the forget scope. Put the key and the site's address in your environment:
export KAIBO_URL=<the address you opened this site at>
export KAIBO_KEY=kaibo_<id>.<secret>
Every client reads KAIBO_URL and KAIBO_KEY on its own, so your code never needs to mention them.

2. Install a client

bun add kaibo        # or: npm install kaibo, pnpm add kaibo, yarn add kaibo
Not sure what your project needs? Run kaibo init in it. It prints the right install command for your package manager and sets up your agent tools (step 4).

3. Remember, change, recall

import { Kaibo } from "kaibo";

const kaibo = new Kaibo(); // reads KAIBO_KEY
await kaibo.remember("The release target is July 13.");
await kaibo.remember("Move the release target from July 13 to July 17.", { wait: true });

const pack = await kaibo.recall("release target");
console.log(pack.found ? pack.answer : `no answer: ${pack.reason}`);
console.log(pack.sources); // the events it came from

4. Let your agent use it

Run kaibo init in your project. It finds the agent tools you already use and writes what they need: a skill for Claude Code, a rule for Cursor, a block in AGENTS.md for Codex and others, and an MCP config. Nothing is overwritten, and your key stays in the environment.
Or paste the setup prompt (/agents/setup-prompt.md on this site) into your agent and let it do the steps. See Use with agents for the details.
To put memory into your own prompts, ask for text:
const memory = await kaibo.context("what did we decide about the release?");
// delimited as stored data, with citations, and with outdated, unverified
// and untrusted items marked. Say "I do not know" when it says there is no answer.

What you get back

The answer is the July 17 memory, cited to the second event. The July 13 memory is still stored, marked superseded, with the time range it was true. Ask recall with asOf set to an earlier time and you get July 13 back.

Check the answer

explain returns the chain: what the memory replaced, and the source events with their hashes. kaibo verify recomputes the hash chain of a stream on your machine.

Forget it

const preview = await kaibo.forget({ query: "release target" }, { reason: "demo", dryRun: true });
const { receipt } = await kaibo.forget({ query: "release target" }, { reason: "demo" });
await kaibo.forget.verify(receipt.id); // { ok: true, ... }
Forgetting needs a key with the forget scope.