Zed 1.8: Git Graph Commands, Fast Mode, and Worktree Flows Zed 1.8.0 shipped June 17 with a custom command context menu on the git graph, cleaner worktree-thread integration, a fast mode toggle for Anthropic and OpenAI models, and a setting to enforce team commit message conventions. The update closes gaps in git UX and improves agent workflows without a headline feature. Zed 1.8.0 shipped June 17 — a focused release that doesn’t try to be a landmark but tightens the gaps developers actually complain about. The update ships a custom command context menu on the git graph, cleaner worktree-thread integration for parallel agent setups, a fast mode toggle for Anthropic and OpenAI models, and a long-overdue way to enforce your team’s commit message conventions. No headline-grabbing feature. Just the editor getting better at what it already does well. The Git Graph Context Menu This is the headline feature, even if Zed didn’t frame it that way. Developers have been dropping to the terminal for operations that require specifying a branch by name — cherry-picks, rebases, targeted diffs — because the visual git graph had no way to act on what it showed you. 1.8 fixes that with a right-click context menu on any branch, remote ref, or tag in the graph. The target ref is injected as $ZED GIT REF , so your custom commands can reference it directly: Cherry-pick the selected branch's tip git cherry-pick $ZED GIT REF Rebase current branch onto what you clicked git rebase $ZED GIT REF Diff against the selected ref git diff main..$ZED GIT REF Define these once and they’re available on every right-click. For developers who’ve been keeping Zed open and a terminal nearby for branch operations, that context switch just disappeared. This is what VS Code users building on GitLens have had for years — and the gap in Zed’s git UX has been a legitimate friction point. 1.8 closes it. Worktree UX: First-Class Now Zed’s parallel agents feature https://zed.dev/docs/ai/parallel-agents launched in April runs agent threads in isolated git worktrees so multiple agents can work the same repository without stepping on each other. The concept was right; the setup was clunky. 1.8 fixes the ergonomics. The most useful change: you can now create a new worktree directly from the sidebar’s new-thread button, rather than setting one up separately before attaching it. When a project has multiple open worktrees, the agent thread picker lets you choose which one to attach to. And you can remove a worktree from the current window without navigating away. There’s also a new agent.terminal init command setting that auto-runs a setup script when an agent opens a new terminal thread. Pair this with worktree creation for fully scripted agent environments — your agent gets a clean branch, runs your setup, and starts working without you managing any of it manually. Fast Mode: Turn It On, Turn It Off Zed’s agent panel now has a fast mode toggle for Anthropic and OpenAI models. On the Anthropic side, this activates priority processing; for OpenAI, it’s their premium service tier. Both deliver lower latency at a higher per-token cost. The practical guidance is simple: use fast mode when you’re interacting with the agent in real time — inline edits, quick fixes, active back-and-forth — and leave it off for long-running background tasks. A background agent rewriting your test suite doesn’t need to respond in under a second. You, mid-coding-flow, do. The toggle exists for exactly that distinction. Enforce Commit Message Conventions AI-generated commit messages have been one of the persistent rough edges of editor-integrated AI: useful enough to use, inconsistent enough to be annoying. The new agent.commit message instructions setting lets you inject any instructions into the AI’s commit message prompt. Add it to your project’s .zed/settings.json : { "agent": { "commit message instructions": "Use Conventional Commits format: