SShortSingh.
Back to feed

Agent Control: Open-Source Remote Console for Managing AI Coding Agents from Your Phone

0
·2 views

A developer has released Agent Control, an open-source, self-hosted desktop and mobile console under the AGPL-3.0 license for monitoring coding agents such as Codex, Claude Code, Antigravity, and Grok. The tool aims to reduce the strain of juggling multiple IDE windows and lets users review diffs, approve or deny tool calls, and queue instructions from a phone. It reads upstream agent histories in read-only mode and stores custom metadata separately, with no cloud storage of prompts or code. Remote access is provided through a QR-code pairing flow over Cloudflare Quick Tunnel by default, with options for Tailscale, Cloudflare Named Tunnel, or a self-hosted relay. The project also includes a hands-free voice mode and a five-language interface.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

PWA or native app: when a web app suits a business better

The article argues that many businesses asking for an app really need a website or a progressive web app, and that building a native app before validating demand is an expensive mistake. A PWA is a website with a manifest and service worker, so it installs to the home screen, works offline, and skips app store review. Rough cost and timeline comparisons put websites at about $500 in 3–10 days, PWAs at about $800 in 2–4 weeks, and native apps at about $1,500 in 4–10 weeks plus store fees. The author lists cases that favor a PWA, such as validating a hypothesis, internal field tools, frequent repeat visits, and tight budgets, and cases that call for native, such as heavy background processing, complex hardware, or store-only requirements. A car dealer example describes a PWA built in 1–2 weeks for vehicle inspections, which confirmed demand and avoided store accounts and moderation.

0
ProgrammingDEV Community ·

Morocco Transport Site Uses Build-Time Checks to Catch Price Errors Before Deploy

A small Moroccan transport booking business, navetteforyou.com, operates across eight languages with more than 640 pages and thousands of price points. To avoid costly pricing mistakes, it built a build-time validation pipeline that crawls every generated page and extracts currency amounts and structured data. Any price below a set minimum fails the build, and cross-language mismatches are flagged as warnings. The team also uses AI to draft translations, generate Schema.org markup, and define the patterns the checker looks for, while the validation itself relies on deterministic matching. According to the author, no pricing errors have reached production since deployment, with about four seconds added to build time.

0
ProgrammingDEV Community ·

Alibaba's Qwen4-Max Launches as Closed API, Breaking Earlier Open-Source Pledge

Alibaba's Qwen team had recently pledged to keep its models open source, but it has now released its flagship Qwen4-Max without downloadable weights, offering access only through a closed QwenCloud API. The model reportedly has 2.4 trillion total parameters, 71 billion active per token, and a 1M-token context window. Its architectural changes reportedly cut token usage on Terminal-Bench 4.0 by about 48%–49% and lower average Work360 task cost by 62%. The author argues that, because teams cannot self-host the model, these efficiency gains do not provide the sovereignty or fallback options that open weights would, creating vendor lock-in risks.

0
ProgrammingDEV Community ·

Recording Product Demo Videos With Playwright: Six Common Failures

Playwright can capture a browser session as video in a few lines of code, which makes automated product demos appear simple to produce. However, the raw recording is only a test artifact and falls short as a demo because it lacks a visible cursor, shows text appearing instantly, and does not wait for the page to finish loading. Its output also looks blurry when zoomed in. The article notes that a demo additionally needs a clear story, defining the task being shown and its audience before scripting begins. It presents fixes for these issues, including logging cursor movement to overlay later, setting device scale at the browser level to get sharper footage, and waiting for page mutations to settle rather than using fixed delays.