Meet Mercury Code. A senior software developer-grade AI agent, built directly inside Mercury Agent. Open Mercury. Type: /code And start building. Mercury Agent v1.2.3 is live ↓

Mercury Code is built for real software work. Give it a task and it can understand your repository, inspect the codebase, plan the implementation, edit files, run tools, verify its work, and keep moving. All from inside Mercury Agent.

No separate coding agent. Just launch Mercury Agent and hit: /code You’re now working with an agent designed specifically around software development.

Mercury Code can handle the full loop: Understand the repo → plan → implement → test → verify → iterate. And when something blocks the task, it can recover, retry, and continue instead of simply stopping at the first failure.

We’ve also put a lot of work into the experience around it. v1.2.3 brings major improvements across coding sessions, execution reliability, context handling, task recovery, verification, and the Mercury Code interface. Plus a long list of fixes across Mercury Agent.

You can follow what Mercury Code is doing while it works: • plans and task progress • file changes • reasoning previews • repository context • verification • blockers and recovery Less black box. More actual engineering workflow.

Mercury Code is available now in Mercury Agent v1.2.3. Install or upgrade Mercury Agent, then type: /code Your senior software developer-grade AI agent is one command away.

GitHub:

The slash command is a smart wedge. People don't need another agent pitch, they need one obvious door into the workflow they already have.

💪

Mercury Code has permission gates. WealthMind Pro has one too: every feature must prove it won't become a new subscription we forget to cancel.

🙌🙌🔥🔥

shipping Mercury Agent v1.2.3 with a plain `/code` entry point is delightfully low-friction; does it also explain the changes before touching an existing project?

A coding agent inside an agent is actually an interesting direction. /code and you get a developer-focused mode that can inspect the repo, make changes, run checks, and iterate. Going to test Mercury Code on a real codebase.

What does 'senior developer-grade' translate to in practice is that anchored to a specific benchmark, or more about autonomy across multi-file tasks?

Okay, /code is definitely the first thing we’re trying in this release.

whats the cost and how does it compare to code from deepseek 4.1

A senior dev hiding behind /code? Yeah, we’re testing this.