cn— for agents who shouldn't forget

Persistent memory
for AI agents.

A shared workspace your agents read and update from the CLI. Structured notes, relationships, and project context that survive the end of a session.

See a live workspace — freeSee CLI docs

Free · one-click Google sign-in · pre-populated with real notes

npx @creator-notes/cli@latest init

▶ 80 seconds: npm install → connected canvas

Works with Claude Code, Cursor, ChatGPT, LangChain, and any agent with CLI or MCP access.

The problem

Agent sessions
don't carry over.

Your agent maps a codebase, learns your preferences, and makes good decisions — then the session ends. Next session, it has no access to any of that work.

So what happens?

  • Re-explain context every session

    repeated setup

  • Same mistakes on already-solved problems

    lost learnings

  • No way to build on prior work

    starting from zero

  • Agents can't share what they know

    isolated context

Markdown files, vector stores, prompt stuffing — all fragile, all maintenance debt.

A workspace
your agents can write to.

CreatorNotes is a knowledge workspace with a CLI and MCP server. Agents create, search, and connect notes through it. Everything persists across sessions and stays visible to humans.

— Structured notes with typed relationships.

— Spatial canvas to see how context connects.

— CLI + MCP. No SDK to learn.

How it works

Three commands.
That's the install.

  1. 01

    Install

    One command sets up everything.

    Auth, workspace selection, and AI tool integration in 30 seconds.

    npx @creator-notes/cli@latest init
  2. 02

    Connect

    Your agents get tools automatically.

    cn init wires up Claude Desktop and Codex via MCP. Custom agents get shell access to cn.

    cn search semantic "auth flow"
  3. 03

    Remember

    Knowledge compounds across sessions.

    One agent’s discoveries become another’s starting context. Humans review and curate in the web UI.

Native integration

Plugs into the agent stack
you already use.

Claude Desktop · Codex

MCP server configured automatically. All 47 CreatorNotes tools available inside your AI chat.

Run cn init and select your client. No JSON editing, no manual config.

Claude Code · Cursor

Agent skill installed to ~/.claude/skills/.

Your coding agent learns the full CLI automatically — every command, every flag, every example.

LangChain · CrewAI · custom

Shell access to cn is enough.

Or set CN_TOKEN + CN_WORKSPACE for headless use. No browser, no SDK.

Example workflow

Two sessions.
One memory.

What it looks like when one coding agent picks up where another left off.

Session 1 — agent documents decisions as it works

# Agent creates a typed note (title comes from the # heading)cn notes create --type Decision --markdown-file ./postgres-decision.md→ Created DECISION-1: "Use Postgres over DynamoDB" # Agent links a related insight via mention syntax in markdowncn notes create --type Insight --markdown-file ./auth-tokens.md→ Created INSIGHT-1: "Auth flow requires refresh tokens"

Session 2 — a different agent finds prior context

cn search semantic "database choice"→ DECISION-1: "Use Postgres over DynamoDB" (score: 0.94) cn notes get DECISION-1→ Full note with content, relationships, and version history

You — review everything in the browser

Open the canvas → decisions, insights, and relationshipsvisualized spatially. Curate, reorganize, add your own context.

Built for people
building with agents.

If you keep watching your agent re-discover the same things — this is for you.

CreatorNotes is for:

  • Agent developers who need persistent context across sessions

  • Teams using Claude Code or Cursor on shared codebases

  • Engineers adding memory to custom LangChain or CrewAI agents

  • Technical PMs tracking decisions and project state via agents

Common questions

How do agents access CreatorNotes?

Through the cn CLI or the built-in MCP server. Any agent that can execute shell commands or connect to MCP tools can create, read, update, and search notes. No special SDK needed.

Is this a vector database?

No. CreatorNotes stores structured notes with titles, content, and named relationships — closer to a knowledge graph than embeddings. Semantic search exists too, but the primary unit is a typed, linked note.

What agents does this work with?

Claude Code, Claude Desktop (via MCP), Cursor, ChatGPT with code interpreter, and custom LangChain or CrewAI agents. If it can run a command or speak MCP, it works.

Is there a free tier?

Yes. Start free with no credit card. The free tier includes full CLI access and enough capacity to evaluate the platform with your agents.

Can multiple agents share the same memory?

Yes. Agents read and write to the same workspace. One agent’s discoveries become another’s context. Humans see and curate everything through the web UI.

Stop re-explaining.
Start remembering.

Explore a live workspace built by agents. Then install the CLI and let your own agents build their own.

See a live workspace — free