Safe, but stalls.
The run blocks whenever the local agent needs a human answer and you are away from the terminal.
/* least-permission approvals */
Your local agent can update you, ask a bounded question, or wait for approval — without giving your phone a shell.


Ask your agent to install Agent Tick with https://agenttick.sh/skill.mdAnimated illustration: a local agent interface and the Agent Tick Native App show the same request. A response from either surface resolves the request everywhere and local work continues.
// mirrored prompts
Your local agent blocks on one Agent Tick Request. Answer in the local agent interface or the Native App; the same bounded response resolves the Mirrored Prompt everywhere, then local work continues.
// bounded, not remote control
Agent Tick is not remote access and not full-auto permission skipping. Your Local Agent Interface creates a bounded request; the Native App or Personal Console returns one allowed response.
The run blocks whenever the local agent needs a human answer and you are away from the terminal.
Remote shells and permission skipping keep work moving by giving the remote surface too much authority.
The local agent asks one bounded thing. The remote surface returns only the allowed response.
Read the security model →// how it works
A status update, a bounded question, or approval for one specific action.
No terminal stream. No arbitrary prompt channel. Just the response the request allowed.
The agent receives the answer and continues locally, or stops safely when you say no.
Status Update: one-way progress, such as waiting, blocked, done, or failed.
Steering: choose from bounded options supplied by the agent.
Sanction: approve or deny one specific action before it happens.
// setup
The setup skill inspects your agent config, explains the changes, asks before editing, and verifies the loop when it is done.
Ask your agent to install Agent Tick with https://agenttick.sh/skill.md Open setup skill// ownership
Inspect the source, run your own server, and try the app before buying. Hosted routing exists if you want convenience; it is not the point of the homepage.
// likely questions
Remote access is the right tool sometimes. Agent Tick is for one smaller job: answer a bounded request without opening a shell to the whole computer.
Because removing the review point is not the same as approval. Agent Tick keeps the review point and moves the answer to a convenient surface.
No. The Native App answers the request shape created by the local agent. It cannot turn into a new prompt channel or remote shell.
Request type, prompt/context needed to decide, allowed responses, selected response, and routing metadata.
The working tree, terminal, commands, agent process, and execution environment stay in the local agent environment.
Yes. Agent Tick is source-available and self-hostable. The hosted service is convenience, not a required black box.
Yes. Mirrored Prompts can resolve from either surface. The first answer resolves the request everywhere.
The local agent receives the denial and should stop or choose a safer path instead of continuing the requested action.
// try it
If it earns a place in your workflow, buy the app once. If it does not, you lost less time than one local-only stalled run.