The human-in-the-loop reviewer for agentic coding.
Agent it. Review it. Feed it back.
Anchor Code gives human reviewers the codebase, technical documents, file management, Git history, diffs, review comments, and agent CLIs in one place. Agents keep working in their native terminal environment while you inspect the context, direct changes, send feedback, and decide whether the result is ready.
Agentic coding has accelerated implementation, but the surrounding human workflow remains fragmented. Code lives in an editor, plans live in Markdown, changes live in Git, feedback lives in chat, and the agent runs in a terminal.
Anchor Code reconnects these surfaces. Browse the codebase and technical documents, run the agents you already use, inspect their actual changes, discuss exact lines, send structured feedback, and decide when the result is ready.
| Code and documentation | Agent CLI sessions | Git and Diff | Comments and Feedback |
|---|---|---|---|
| Browse files, search code, and read source, Markdown, and Mermaid documents. | Run native Agent CLI and TUI sessions in the current project, with model selection and reasoning settings. | Review commit history, branch and worktree state, and complete changes in side-by-side or inline diffs. | Comment on exact code selections, send review sessions to agents, and verify replies and revisions in the original context. |
Select any two commits, or compare a commit with the current worktree. Anchor Code presents the complete changed-file set and diff in one focused workspace, with side-by-side and inline viewing modes.
Branch state, revisions, files, source lines, and comments stay together. A comment created in a diff retains its branch, base, head, file, and line-range context, giving both the human and the agent an exact reference.
Comments are anchored to exact selections in source code and rendered Markdown. Review sessions distinguish questions (discussing), concrete change requests (need_modify), and resolved threads (closed).
The Feedback action sends the structured session to the selected agent. The agent can reply in the original thread, implement requested changes, and update its status. The next diff and the full discussion remain available for human verification.
Zed and VS Code center the editing experience. Warp centers the terminal and agent experience. Anchor Code connects files, documents, Git inspection, review context, and Agent CLIs around a continuous human-in-the-loop workflow.
✅ built in · △ available through a broader workflow, integration, or extension · ❌ no comparable workflow
| Capability | Anchor Code | Zed | VS Code | Warp |
|---|---|---|---|---|
| Files, code, and Markdown | ✅ Built in | ✅ Built in | ✅ Built in | △ Terminal-oriented |
| Integrated terminal and Agent CLI | ✅ Native CLI/TUI | ✅ Integrated terminal | ✅ Integrated terminal | ✅ Core experience |
| Git commit history | ✅ Review-oriented history | ✅ Project/file history | ✅ Source Control Graph | △ CLI and integrations |
| Commit ↔ Commit / Worktree comparison | ✅ First-class Diff workspace | △ Commit and file diffs | △ Built-in Git views; GitLens adds deeper comparison | △ Primarily CLI-driven |
| Comments anchored to code and diffs | ✅ Review Sessions | △ Editor/collaboration workflows | △ Extensions or PR integrations | ❌ |
| Structured feedback returned to CLI agents | ✅ Stateful feedback loop | △ Agent/editor context | △ Depends on Agent extensions | △ Terminal conversation |
The difference is what happens around the diff. Anchor Code makes comparison, code-anchored discussion, Agent handoff, and human verification one native path.
Comparison references: Zed Git, VS Code Source Control, GitLens features, and Warp documentation.
Anchor Mobile extends the workspace to Android phones and tablets. Review code and Markdown, inspect worktree diffs, manage comments, and operate Agent terminal sessions running on your PC.
Pair with a QR code through the end-to-end encrypted Anchor Relay. Files, Git operations, shells, and agents remain on the PC; the Relay forwards encrypted frames. See Anchor Mobile for setup and APK instructions, and Anchor Relay for deployment details.
Agent CLIs: Claude Code, Codex, Gemini CLI, Aider, Grok, OMP, Cursor Agent, and custom profiles.
Workspaces: Local, WSL, and SSH.
Desktop: Windows, macOS, and Linux.
Companion: Android phones and tablets.
- Download Anchor Code from Releases.
- Open a Local, WSL, or SSH workspace.
- Launch an Agent CLI in the built-in terminal, or open a repository with existing changes.
- Read the code and documentation, then use History to inspect commits or the worktree.
- Add comments to exact selections and mark actionable requests as
need_modify. - Use Feedback to hand the session to an agent, then inspect the next diff and close resolved threads.
Install the Anchor Review skill from Settings → Agent skill, or accept the workspace prompt. It teaches compatible agents how to read Anchor Code sessions, process need_modify comments, reply to threads, and update review status.
| Platform | Installer | Portable / other |
|---|---|---|
| Windows | Anchor.Code-*-win-x64.exe (NSIS setup) |
Anchor.Code-*-win-x64-portable.exe |
| macOS | Anchor.Code-*-mac-*.dmg |
.zip |
| Linux | .AppImage or .deb |
- |
Anchor Code requires system git on the selected host. WSL and SSH workspaces require their corresponding host environment and connection to be available.