Show HN: Pulse – Dashboard for Claude Code, approve tool calls from your phone

What Happened

A student developer from Flanders created "Pulse," a local dashboard application that enables remote approval of Claude Code tool calls from a mobile phone. The tool addresses a practical friction point in the Claude Code workflow: when Claude autonomously requests permission to execute actions (like file edits, shell commands, or API calls), users previously had to be physically present at their development machine. Pulse bridges this gap by routing approval requests to a phone interface, allowing asynchronous oversight of Claude's actions.

The project is self-described as a local application, suggesting it runs on the user's own infrastructure rather than relying on cloud services. This design choice prioritizes privacy and latency, as tool call approvals happen without sending data to third-party servers.

Why It Matters

This development highlights a growing tension in AI-assisted coding: the desire for autonomy versus the need for human oversight. Claude Code already operates with a "human-in-the-loop" model for sensitive operations, but that loop was tethered to the desktop. Pulse effectively extends that loop to mobile, making the approval process more flexible.

For practitioners, this addresses a real-world pain point. Developers often run long Claude Code sessions where the AI hits a permission boundary while the user is away from their desk—grabbing coffee, in a meeting, or working on another task. Without remote approval, the session stalls or the user must pre-authorize broader permissions, which defeats the safety purpose of tool call approval. Pulse offers a middle ground: maintain granular control without being chained to the workstation.

The fact that a student built this independently also signals a broader trend. The Claude Code ecosystem is maturing to the point where third-party tooling can meaningfully extend its capabilities. We're seeing the emergence of a community-driven layer around Claude Code, similar to how early plugins emerged for VS Code or how community scripts enhanced AutoGPT.

Implications for AI Practitioners

Workflow flexibility: For developers who run Claude Code as a background assistant or in long-running sessions, Pulse could reduce context-switching friction. Instead of interrupting flow to approve actions, they can handle approvals from their phone during natural breaks. Security considerations: While the local-only design is prudent, any remote approval system introduces new attack surface. If the phone interface is compromised, an attacker could approve malicious tool calls. Practitioners should evaluate whether Pulse encrypts the communication channel between phone and desktop, and how it authenticates the user. Scaling human oversight: This model could extend beyond individual developers. Teams running shared Claude Code instances (e.g., on a build server) might use similar dashboards to distribute approval responsibilities across team members, with each person approving actions relevant to their domain. Ecosystem maturation: Pulse represents a shift from "can Claude code?" to "how do we safely integrate Claude into existing workflows?" The next wave of Claude Code tooling will likely focus on governance, audit trails, and approval routing—areas where enterprise adoption will demand robust solutions.

Key Takeaways

• Pulse solves a concrete usability gap by enabling mobile approval of Claude Code tool calls, freeing developers from desk-bound oversight.
• The local-only architecture aligns with security best practices but introduces new considerations around phone-to-desktop authentication and channel encryption.
• This tool signals the emergence of a third-party ecosystem around Claude Code, with community developers building infrastructure for safer, more flexible AI-assisted coding workflows.
• For practitioners, Pulse demonstrates that human-in-the-loop systems can be made more practical without sacrificing granular control—a critical balance for production use cases.