MCP server: One MCP server.

Sticky

@1picassoaiRun by its publisher: served from their own domain, namespace checked by the registry.MCP server · External

Persistent memory and context for your agent, on a board you can see. Capped, so it stays small.

AI memoryv0.2.2JavaScript
Freeno charge per call

Runs on your machine from its package; the steps are below.

image
From the README

What it does

  1. Reads
  2. Writes
  3. High impact

At most: creates or changes data: records, files, settings.

What it costs

Free

Odel charges nothing per call. The publisher may charge for its own service.

Runs locally over stdio

This server isn't hosted — your MCP client launches it from a package registry. Use one of the commands below, or drop the config into your client (e.g. Claude Desktop).

npmsticky-mcpv0.2.2

Run

npx -y sticky-mcp

MCP client config

{
  "mcpServers": {
    "sticky": {
      "command": "npx",
      "args": [
        "-y",
        "sticky-mcp"
      ]
    }
  }
}
README3 min read
image

A board your coding agent can write to — and you can see.

Your agent forgets between sessions. The usual fix is a CLAUDE.md that grows until it's a thousand lines of stale rules nobody deletes, quietly costing you tokens on every request.

STICKY holds ten active notes. When the eleventh arrives, the oldest one tumbles to the archive on its own. Nothing to prune, nothing to garden.

bash
npx sticky-mcp

That's it. The board opens in your browser. No account, no cloud, no config.


Wire it to your agent

bash
claude mcp add sticky -- npx sticky-mcp-server

Cursor, Cline, Windsurf and anything else that speaks MCP work the same way — point them at npx sticky-mcp-server.

Or add it to your MCP config by hand:

json
{
  "mcpServers": {
    "sticky": {
      "command": "npx",
      "args": ["-y", "sticky-mcp-server"]
    }
  }
}

sticky-mcp-server takes no flags on purpose. Some CLIs — Claude Code among them — read -y and --mcp as their own options and never pass them through, so a command with nothing after it is the only one that works everywhere.

Your agent gets two tools:

  • post_sticky — pin a decision, a constraint, or where the work paused
  • read_active_stickies — read the board at the start of a session

The three columns

what it holdsreaches the agent
Pinnedpermanent rules — "use C# 12", "never touch migrations by hand"always
Activecurrent decisions and where work stoppedalways, newest 10
Archivewhatever tumbled offnever — zero tokens

Drag a note between columns. Drop it on Pinned and it becomes a permanent rule. Drop it on Archive and it stops reaching the agent immediately. Double-click to edit, click the cross to bin a bad assumption.

The board updates live as your agent posts — you watch it think.

What's bounded, and what isn't

Being precise, because a promise that isn't quite true is worse than no promise:

  • Active is capped at 10 (configurable 5–20). The oldest unpinned note tumbles out.
  • Pinned is capped at 15. Rules never tumble — a rule vanishing silently would be worse than the bloat — so instead the board refuses the sixteenth and asks you to retire one.
  • Archive is unbounded and costs nothing, because it never reaches the model.

So the agent's context stays bounded at roughly 25 notes, permanently, no matter how long you work.

Your data

The board is a SQLite file at ~/.sticky/store.db. The web server binds to 127.0.0.1 and nothing else. There is no telemetry, no analytics, and no network code in this package beyond the local server itself.

Options

npx sticky-mcp                 open the board
npx sticky-mcp --mcp           run as an MCP server (agents launch this)

  --cap <n>      active notes before the oldest tumbles (default 10)
  --port <n>     board port (default 7317)
  --db <path>    where the board lives (default ~/.sticky/store.db)
  --no-open      start without opening a browser

Want it floating beside your editor? chrome --app=http://localhost:7317

Requires

Node 22.5 or newer — STICKY uses Node's built-in SQLite, so there's nothing to compile and no native modules to install.


MIT licensed. Built by Juvina.

Similar modules

Browse AI memory