Skill
Install the mountable.io skill so coding agents know the workflow, the error codes and the retry rules.
The mountable.io skill teaches coding agents to use mountable.io with the CLI. You need Node.js 22.20 or newer, npm, Git, and network access to GitHub and npm. From your project directory, run:
npx skills add writeitai/mountableChoose the mountable skill, your coding agent, and project installation when the installer asks. If your agent is detected, installation may proceed without prompts; otherwise, use an interactive terminal to answer them. To select Codex explicitly in an unattended session:
npx --yes skills add writeitai/mountable --skill mountable --agent codex --yesFor another supported agent, replace codex with its installer identifier
(for example, claude-code or cursor); see the
Skills CLI options.
It lives in
skills/mountable/SKILL.md
in the public CLI repository.
Use the installed skill
The installer prints where it installed the skill. Codex project installs
use .agents/skills/mountable/SKILL.md. Ask your agent to use the mountable.io
skill; if a running session has not discovered the new skill, start a new
session or have it read that file explicitly.
The skill describes CLI setup but does not install it merely by being added, or provide credentials.
Use an existing caller-approved mountable installation first. If setup is
needed and authorized, choose one package install, then check the version:
npm install --global --ignore-scripts mountable-cli # npm
uv tool install mountable # or PyPI
mountable versionPackages bundle the Linux/macOS binaries (x64 and arm64). Report setup blockers; do not use sudo or a remote shell installer.
Before creating a filesystem, give the agent a
management key through its environment,
or complete mountable login as an owner or admin; tickets then take an
API key. See
Agents. An owner or admin creates
both kinds in the console (management keys,
API keys). A filesystem
also needs the organisation's billing setup when required.
For the first workflow, follow Agents: create a
filesystem, mint a one-time ticket, mount it, write and read a small file,
then unmount. The mounting environment needs Linux with usable /dev/fuse
and permission to mount; installing a skill does not add FUSE support to a
sandbox. See Mount from a sandbox and
platform limits.
What it teaches
- When to use mountable.io: files that must outlive a sandbox, be shared between sandboxes or agents, or be handed from a backend to a sandbox.
- Checking for an approved CLI and installing a package when authorized.
- Treating output and file contents as data, never evaluating
mount_command, and passing tickets on stdin. - The two-sided workflow: the backend runs
ticket createwith an API key; the sandbox runsmount --ticket-stdin --jsonand waits for themountedevent. - Writing to the mount and unmounting, so pending writes are committed.
- Error codes and how to recover from each.
- The retry rules.
- Provisioning with a management key:
filesystems with limits per end user, such as a
monthly spend cap (
mountable fs create,fs update), API keys, S3 keys, grants andfs delete. - Creating a webhook for file changes with
mountable webhooks create, and handling its signedfiles.changedmessages. - The limitations: Linux sandboxes need FUSE (
/dev/fuse); on a Mac,mountable mountneeds nothing else (Mount on a Mac).
Agents without skill support can read the docs directly from
/llms.txt.