Features

Everything your
library can do.

Start my library

The library

Plain files you own, at real paths, that a person and a model read the same way.

  • Notes, prompts, attachments and folders, kept as plain files at real paths
  • Skills: a folder holding SKILL.md and its supporting files
  • Turn a note into a prompt, or a folder into a skill, and back
  • Full-text and filename search across the library
  • Tags across items, with a page for each tag
  • Archive and restore anything; nothing you can reach is destroyed
  • Drag a whole directory in from the desktop, folder shape kept
  • Mirror a GitHub repo, or part of one, into a folder, pinned to a branch, tag or commit, and kept up to date overnight
  • Links between files resolve the same way in the browser, in an editor and for a model
  • Items generated from the body of work you already have

Reach

One library, reached from every AI app and every device.

  • Connect any AI app over MCP: one connector URL, one consent screen
  • Guided setup for Claude, ChatGPT, Cursor, VS Code and Codex CLI, and a path for any other app that speaks the standard
  • Seventeen tools a model can call: list, read, search, follow links, create, write a version, move, archive
  • One connection reaches every library you belong to
  • See and revoke connected apps at any time
  • The same library at your desk, on your laptop and on your phone
  • A REST API with a published OpenAPI spec

Versions and control

Nothing a model writes sticks until you say so.

  • Every file keeps its full version history, with who or what wrote each version
  • Diff any two versions, and roll back in one click; a rollback never destroys anything
  • Anything a model wrote stays marked until you accept or roll it back, whether it came from a cloud run or from your own AI app
  • One model pass makes one version, not a hundred
  • How each file came to be: the step, the model, the turns, the tokens and the cost Cloud

Team

One shared library, and a personal one for every member.

  • A shared library the whole team reads, writes and runs against
  • Members, roles and email invitations
  • Every member has a personal library of their own, alongside the team's
  • Switch between your libraries from the header
  • Transfer a library to another owner
  • A team dashboard: the library's home and its usage stats
  • Read-only guest access for a coach or consultant: one folder, plus a private dropbox they can write into
  • Nothing is public, and nothing is shared unless it is in a shared library

Stats

See what you and your team actually use.

  • Library stats: what your connected AI apps did this month, by file, by skill, by member and by app, with a daily chart
  • Compute stats: hours used against the monthly allowance, by member and by workspace Cloud
  • Spend stats: AI spend by provider, member, workspace and credential, with your own keys shown as not billed Cloud
  • Credit balance, with low-credit and near-limit warnings Cloud
  • A checklist that walks you from first connection to first run

Ownership and export

Plain markdown, yours to take.

  • Export the whole library in one click, as a ZIP of plain markdown with skills as folders
  • Download any note as markdown, any folder or skill as a ZIP, any attachment as its bytes
  • Download a whole workspace version, or a single file from it Cloud
  • Per-file licenses, inherited from the nearest folder
  • Stop paying and the library goes read-only on every surface, and stays exportable
  • Delete your account and everything in it

Cloud Cloud

Pro Cloud and Team Cloud run the work on our machines instead of yours.

  • Work runs off your machine, in a sandbox, and finishes with your laptop closed
  • Several runs at once
  • Cloud workspaces: a set of files a run builds in, for file- and data-heavy work
  • Browse, edit, upload and delete workspace files in the browser
  • Workspace versions with history, diff and download, to accept or reject
  • Your own working draft on a version: save, promote, reject or restore
  • Open a GitHub pull request from a workspace version
  • Import a workspace from a ZIP, or sync it from a GitHub repo
  • 50 hours of compute a month per seat

Runs Cloud

Point a model at your files and it does the work there, not in a chat.

  • Point a run at a note or a folder; the model reads the library, does the work and writes the results back
  • Queue several prompts as steps and reorder them before running
  • Watch a run stream live; stop, resume or continue it
  • One diff showing everything a run changed across the library
  • Recipes: save a sequence of steps and run it again on another note or folder
  • Long runs can pause themselves and resume later on their own
  • Every run gets a name and an emblem, so a finished piece of work is something you can find again
  • Every task starts with a fresh context: the files you chose and nothing else

Models and credentials Cloud

Any model, on your subscription, your keys or our credits.

  • Connect your Claude or ChatGPT subscription, or bring your own API keys
  • Buy credits and run on our keys instead
  • Choose the provider, model and reasoning effort per task, or run the defaults
  • An integrations hub listing AI providers, git hosts and services
  • A skill can declare the integrations it needs; a step attaches the credentials to meet them

The best AI is Actual Intelligence.

Capture it. Refine it. Reuse it. Share it.

Start my library