@agentapprove/pi
Agent Approve extension for Pi - approve or deny AI agent tool calls from your iPhone and Apple Watch
Install
pi install npm:@agentapprove/pi
Installs from npm into ~/.pi/agent/.
Add -l for project-local (.pi/). Manage with pi list / pi update / pi remove.
What this package can reach
Detected automatically from the source at
version 0.1.12.
This describes access, not safety — you decide whether it's appropriate for what
the package claims to do.
- Behavior — Changes how the agent thinks Modifies the system prompt, injects messages, or rewrites conversation context.
- Commands — Runs shell commands Spawns processes on your machine.
- Credentials — Handles your API keys Reads or passes provider credentials. Highest-trust capability there is.
- Files — Reads files outside your project Touches paths beyond the working directory, such as your home folder.
- Network — Talks to the internet Makes outbound network requests. Check the destinations are ones you expect.
- UI — Draws its own interface Renders widgets, footers, dialogs or full-screen views.
What it actually did when run
observedLoaded in a sandbox with no network, a read-only filesystem and a recording stand-in for pi's API. This is what it called — not what the source suggests it might.
Events subscribed agent_endbefore_agent_startmessage_endmessage_updatesession_compactsession_shutdownsession_starttool_calltool_execution_endtool_result
10 calls the source scan did not predict
on("agent_end") · on("before_agent_start") · on("message_end") · on("message_update") · on("session_compact") · on("session_shutdown") · on("session_start") · on("tool_call") · on("tool_execution_end") · on("tool_result")
Reached through a value the scan could not follow — passed as an argument, stored on another object, or returned by a helper.
Run against version 0.1.12 on 2026-08-01.
Registration-time behaviour under default settings; other configurations may differ.
Reviews
0 publishedSign in with GitHub to write a review, reply, or vote.
No reviews yet — be the first.
More in Guards & policy
- @spences10/pi-nopeek — Secret-safe workflow reminder that steers Pi agents to use nopeek instead of exposing .env…
- pi-mono-sentinel — Pi extension that guards against content-based secret leaks and indirect script execution
- pi-mono-context-guard — Pi extension that guards context window growth by auto-limiting read and rg output
- @diegopetrucci/pi-permission-gate — A pi extension that prompts before dangerous bash commands and protected file writes.