Unstoppable Domains

Unstoppable is showcasing Unstoppable Code's ability to run Claude and Codex side by side in one project — for example, Claude on the checkout flow while Codex handles a nav bug. Each task gets its own git worktree, branch, and working directory, so agents don't end up fighting over the same branch.

• Positions Unstoppable Code as pairing frontier coding models on separate tracks within a single codebase
• Continues the product's push on multi-agent workflows alongside recent model additions and agent tooling

Source: X post

Unstoppable announced that Unstoppable Code can now be driven from the browser as well as the desktop app: users describe what they want built, the platform sets up the workspace, starts the work in the browser, and keeps everything ready to continue in the desktop app. The team also said its web agents now support MCP servers right in the browser, extending agent tooling to the cloud workflow.

• Cloud workspaces let users switch between browser and desktop without losing progress
• MCP server support arrives for web agents in the browser

Source: X post

Developer Dani Avila spotlighted Unstoppable's agentic coding IDE for letting users run Claude and Codex together in code review loops, with customizable workflows. Unstoppable retweeted the post on its official X account, showcasing the IDE's cross-model agent collaboration.

• Positions the IDE as able to pair frontier coding models rather than lock users into one
• Comes as Unstoppable Code continues to expand model support and agent tooling

Source: X post

Unstoppable shipped a usage dashboard for Unstoppable Code that tracks token usage, spend, and activity across every session, model, and agent. The view shows chat sessions and completed turns, pipeline runs, and which repos an agent touched, answering the two questions that come up after a day of running coding agents: what actually ran, and what did it cost.

Unstoppable added Claude Opus 5.5 and the GPT 6 family to its AI Credits lineup, joining the models already offered in Unstoppable Code. Users can bring their own subscription or use the free AI credits to stretch their token budget, per the announcement.

The release is framed as keeping pace with the frontier, with Opus 5.5 described as a half-step version jump now available alongside the GPT 6 models.

Unstoppable Domains brought Unstoppable Code to Windows, announcing on its X account that the platform ships with the same app, the same parallel agents, and the same Cowork experience as on other systems. The release is framed as a full port rather than a reduced build, extending the agentic coding desktop app beyond its earlier platforms.

Unstoppable highlighted a round of features shipped in its desktop app over recent releases, per a post by Brady (@ntropiq) reposted by Unstoppable.ai. Users can now import their Claude and Codex chats and projects, share links with one-click run for skills, pipelines, and automations, and optionally queue or steer a running agent with messages.

• A new onboarding flow rounds out the release
• The post is a recap of accumulated desktop app updates rather than a single feature drop

Unstoppable Domains separated its two platforms: unstoppabledomains.com is back as a standalone site for its domain registrar and marketplace, while a separate site now hosts its agentic workspace (Unstoppable Code).

• The registrar/marketplace keeps domain sales, transfers, backorders, auctions, and listing tools
• The agentic workspace gets its own dedicated platform for agents, worktrees, and team usage

Per the company, the split keeps a clear separation of functionality and a high-quality experience for each audience.

Unstoppable Domains added Schedule Resume to Unstoppable Code, letting a chat pick itself back up after hitting a usage limit. When a limit interrupts a task mid-way, the conversation resumes automatically in the same session once the limit resets, continuing where it left off.

• No need to restart the task or re-enter context after a reset

Unstoppable Code shipped automatic model routing, letting the Jev agent pick the best model for every message by weighing required effort level, switching costs, prompt cache impact, and similar factors. The feature removes the need to set a model per chat or per pipeline step.

Announced by Aaron Quirk, who said backtesting against his last few weeks of work produced impressive results.