/
Menu

AI agents

What a coding agent can and cannot do in StackHangar. An agent can ask you for keys and propose playbooks. It never receives a value.

Coding agents such as Claude Code and Cursor can connect to StackHangar and ask you for things. You answer each request in your browser. The agent never receives a value.

#What an agent can do

An agent can:

  • Ask you to store a new key. You paste the value into StackHangar yourself. The agent gets a reference to the saved item.
  • Ask to use a key you already stored. You pick which one and see it in your own browser. The agent only learns that you approved.
  • Ask you to put a secret into an environment variable at a deploy target. See Secret requests.
  • Propose a playbook on one item: rotate a key, or respond to a leaked key. See Playbook proposals.
  • Look up which projects it may ask about, as opaque IDs.

#What an agent cannot do

An agent cannot:

  • Read a value, an item name or a project name.
  • List what is in your vault.
  • Create, rename or delete projects or items.
  • Propose shutting a product down.
  • Ask about a project you didn't give it.

If an agent asks for a project outside its list, the request is refused before it reaches you.

#You approve every request

Each request gives the agent a one-time approval link that works for 10 minutes. The agent shows it to you. You open it on app.stackhangar.com, see which agent is asking and why, and approve or deny. Approving needs a passkey check. Denying doesn't.

The approval happens on StackHangar's own address, never inside the agent's chat. Passkeys only work on the real address, so a look-alike page can't unlock your vault.

#Limits on requests

  • An agent can have at most 5 requests waiting for you, and can make 30 requests an hour.
  • An agent's description of what it needs is shown to you as plain text. If it looks like it contains a key, token or password, the request is refused.
  • Once you answer a request, the agent's description is deleted.
  • StackHangar's agent connection refuses requests that come from web pages, so a site you visit can't use your agent's connection.

#Plans

Free includes 1 connected agent, and that agent can propose playbooks. Solo has unlimited agents and lets them run some playbook steps you allow. See Plans and limits.

#What StackHangar can't control

StackHangar keeps values out of the agent's tools. It can't control what happens after you take a value out of the vault.

Warning

Once you paste a value where the agent can read it, the agent has it. That includes a .env file in the project, a terminal the agent uses, and the agent's chat. An agent that can run any command can read whatever your user account can read. Use the agent's own permission settings to limit that.

If an agent created a key itself, or read one from a file, it saw the value. Treat that key as exposed. See Manage connected agents for how to mark such a key.

#In this section

In this section