Monitor AI Agent Model Routing Performance from Your iPhone

Running AI agent benchmarks, like those for Weave Router 2.0, often means long waits and constant monitoring of terminal output. You need to see performance metrics and task completion in real-time, but staying glued to your Mac isn’t always practical.

A few months ago we started building a model router for coding agents because we thought we could outperform any single model with an ensemble approach. Recently we’ve achieved that milestone and I want to talk about how we did it.

First of all, a quick explanation: the Weave Router (https://github.com/weave-os/router) plugs into any coding agent (e.g. Claude Code or Codex) and intelligently switches between LLMs. So, for example, Astra handles tricky debugging or complex system design tasks, and Deepseek v4 Flash handles simple frontend updates.

What we’re announcing today is our new routing model, which we’re calling Weave Router 2.0. We benchmarked 2.0 against GPT-6 Astra on Terminal Bench 4.0 and SWE Atlas. On both benchmarks, the router had equivalent pass rates. On Terminal Bench, the router hit 52% of Astra’s cost, and completed tasks 2.2x faster. On SWE Atlas, the router cost 54% as much as Astra and ran 2.5x faster. (Full results on our website at https://weaveos.com/router!)

It turns out traini

TermBridge addresses this by mirroring your Mac’s terminal directly to your iPhone or iPad. You can watch the live output of your Weave Router 2.0 benchmarks, agent logs, and performance metrics from anywhere in your home or office. It simply displays your existing iTerm2, Terminal.app, or integrated VS Code terminal, keeping you connected without needing to sit at your desk.

Monitoring Weave Router 2.0 Benchmarks from Anywhere

The Weave Router 2.0 announcement on Hacker News, found at https://news.ycombinator.com/item?id=49911500, highlights significant performance gains and cost reductions for coding agents. Benchmarking these complex systems, especially across Terminal Bench 4.0 and SWE Atlas, can be a time-consuming process. TermBridge allows you to keep a constant watch on these long-running tests. You can see the real-time progress, pass rates, and resource usage directly on your phone, ensuring you never miss a critical update or a completed task, even when you step away from your Mac.

Managing Multiple Agent Runs Across Macs

For developers working with multiple AI agents or running Weave Router 2.0 across different Mac machines, TermBridge simplifies management. You can easily switch between the terminal sessions of various Macs from a single iPhone or iPad. This is particularly useful when comparing benchmark results or overseeing different agent configurations simultaneously. Whether you are monitoring a Deepseek v4 Flash run on one Mac and an Astra debugging session on another, TermBridge provides a unified view. This capability extends to managing various agent workspaces, as discussed in our post on Manage Herdr Agent Workspaces Across Macs from iPad.

Interacting with Agents and Reviewing Logs Remotely

While Weave Router 2.0 aims for autonomous model switching, coding agents sometimes require human intervention or prompt adjustments. TermBridge lets you type commands and respond to agent prompts directly from your mobile device. This means you can review agent logs, debug intermediate failures, or issue new instructions without needing to be physically present at your Mac. The end-to-end encrypted connection ensures your interactions remain private and secure, whether you are checking on Astra’s complex system design tasks or Deepseek v4 Flash’s simple frontend updates.

Discussed on Hacker News · 115 pts · 39 💬