With AI, Everyone is a Dev
Checking profile…
Fresh AI dev tools
View all new tools →Devin Bots
Autonomous Agents for Pull Requests
Devin Bots are goal-driven agents that lead teams of parallel Devin sessions to build, test, and ship software changes through to merge-ready pull requests.
Quicksand
Python QEMU VM Sandbox for Agents
An async Python API to launch, control, and snapshot QEMU virtual machines, focused on sandboxing AI agents.
With Coral
Self Hosted MCP Agent Gateway
Open-source, self-hosted execution layer that gives AI agents policy-controlled access to enterprise systems through MCP and reusable custom tools.
CodeFRAME
Regression Ledger for Coding Agents
Regression memory for coding agents that records fixed bugs as requirements, locks them to affected code, and checks them before every commit.
This Week’s AI Dev Gems
Handpicked tools for the week of Nov 14–20
Write repo automations like issue triage, CI failure analysis and docs updates as Markdown, and run them as GitHub Actions with Copilot, Claude Code, Codex or Gemini doing the work.
Why: Added Nov 18. Useful for chores that need judgment rather than scripts. The agent runs read-only in a sandbox, and its proposed writes are scanned for threats and applied by a separate job with scoped permissions. GitHub Next's project went GA Oct 28 on Free, Team and Enterprise plans.
A self-hosted workspace that runs Claude Code, Codex and eight other agent CLIs through repeatable delivery steps, where tests pass or fail by exit code and humans approve before anything ships.
Why: Added Nov 20. Built for teams that want agent work to be repeatable: separate planner, coder, reviewer and tester roles, Docker sandboxes, and failed checks sent back with the error log. Each run gets a 0 to 100 score with suggested process fixes. Tencent shipped v0.4.0 Nov 17; 5,200+ stars.
Mistral's coding and work agent, now a desktop app for macOS and Windows that runs several coding tasks in parallel, locally or in a cloud sandbox, with diffs and PRs reviewed in the same window.
Why: Added Nov 18. Vibe Desktop went GA Nov 17. Git worktrees keep parallel tasks in one repo from colliding, and dev servers keep running while the agent works. It is a European-made alternative to the Claude and Codex apps, with Pro at $14.99 a month and on-premises deployment for enterprises.
UI behaviors like popups, selection and keyboard focus that your coding agent composes into new components, which then expose their actions over WebMCP so browser agents operate them directly, without screenshots.
Why: Added Nov 17. Agents get building blocks, not fixed widgets, so one set covers a select, a combobox or something new, in React, Vue, Svelte, Solid or Web Components. Vercel Labs says agents using the controls directly are over twice as fast at under half the cost of screenshot clicking. Beta; 6,200+ stars.
AI Dev News
92+ articlesIdeas from the community
Compare Tools
Ask for a solution, not another list.
“I need a self-hosted browser agent that can work with my existing stack. What should I look at?”
Two models search the catalog
Explore their recommendations and reasoning.
You choose the useful answer
Compare perspectives, then investigate the tools.
What would you let an agent do when a production alert fires?
Investigation, repair and deployment can have different approval rules. Choose the furthest step you would allow without human approval.
Keep learning while you build
Explore developer podcasts and the full AI stack guide between deep dives.



