How does Atlas ensure I approve changes before they are written?

Atlas computes a unified diff for every file edit it proposes. This diff is then surfaced in your terminal for explicit approval. Atlas will not write any changes to your files until you have reviewed and approved them, ensuring you maintain full control over your codebase.

Can I reject Atlas's proposed changes?

Yes, you have full control. When Atlas surfaces a unified diff for approval, you can choose to reject the proposed changes. This prevents Atlas from writing the file and allows you to guide the agent towards a different solution or take over the edit yourself.

Where in Atlas's code is this diff review implemented?

The core logic for computing and surfacing diff-reviewed edits is found in the Atlas-Backend. Specifically, it resides in the `atlas/src/tool/edit.ts` file, managed by the `EditTool` symbol.

Does Atlas review changes to new files?

Atlas computes a unified diff for every file edit, which includes the creation of new files (represented as an addition of all lines). This ensures that even new files proposed by Atlas are subject to your review and approval before being written to your project.

Is this feature available in the free core of Atlas?

Yes, Diff-reviewed edits are a fundamental safety and control mechanism within Atlas. As a proprietary tool with a FREE CORE, this capability is available to all users, ensuring a secure and controlled AI coding experience from the start.

How does this integrate with my existing git workflow?

Atlas reads git branches, status, and diffs, and can stage and create commits on your behalf. By requiring approval for every file edit, Diff-reviewed edits ensure that only approved changes are introduced, which you can then review again with standard git tools before committing, or let Atlas stage and commit for you.

Diff-reviewed edits in Atlas (2026)

Diff-reviewed edits in Atlas provide a critical safety mechanism by computing a unified diff for every proposed file modification and presenting it to the developer for explicit approval before any changes are written to disk. This ensures that developers maintain full control over the AI's output, preventing unintended alterations and integrating direct into existing code review workflows in 2026.

What are Diff-reviewed edits in Atlas?

In 2026, Atlas's Diff-reviewed edits capability ensures that every file modification proposed by the AI agent is transparently presented to the developer. This core safety feature computes a unified diff, allowing for a clear, line-by-line review of all changes before they are committed.

Diff-reviewed edits in Atlas are a fundamental safety feature designed to give developers complete oversight of the AI's actions. Before Atlas writes any changes to your codebase, it first computes a unified diff for every file edit. This diff is then surfaced directly within the terminal-native TUI, allowing you to inspect precisely what Atlas intends to change. This mechanism prevents unexpected modifications and ensures that all AI-generated code aligns with your intentions and coding standards. It's a crucial step in maintaining control and trust when working with an AI coding agent.

How Atlas computes and presents file changes for review.

Atlas ensures every file edit is reviewed by first drafting a plan in a read-only agent, then switching to a build agent to compute the changes. This process, in 2026, guarantees that a unified diff is generated and presented for approval before any write operation occurs.

The mechanism behind Atlas's diff-reviewed edits involves a structured agent workflow. Initially, Atlas drafts a plan in a read-only plan agent, ensuring no changes are made prematurely. Once a plan is settled, Atlas switches to a build agent. During the execution phase, for every file edit that the build agent proposes, Atlas computes a unified diff. This diff highlights additions, deletions, and modifications in a standard format. The OpenTUI renderer, powered by SolidJS, then surfaces this diff directly in your terminal, providing a clear, interactive view of the proposed changes, awaiting your explicit approval before proceeding to write the file.

User interaction with Diff-reviewed edits in Atlas sessions.

When working with Atlas in 2026, developers interact with diff-reviewed edits through a clear, terminal-native interface. After Atlas proposes a change, a unified diff is displayed, requiring explicit approval before the file is written, giving you full control over the agent's output.

In a typical Atlas session, after the AI agent has formulated a solution and is ready to apply file modifications, the Diff-reviewed edits capability activates. The terminal-native TUI, rendered with SolidJS through the OpenTUI renderer, presents the unified diff for each affected file. You can review the proposed changes line by line, understanding exactly what Atlas intends to do. Atlas waits for your explicit approval before writing the file. This interactive approval step is integral to the workflow, allowing you to accept, reject, or request further modifications, ensuring that the AI's actions are always aligned with your development process. This also complements Atlas's ability to snapshot file changes as git patches, allowing for easy rollback if needed.

Frequently asked questions

How does Atlas ensure I approve changes before they are written?: Atlas computes a unified diff for every file edit it proposes. This diff is then surfaced in your terminal for explicit approval. Atlas will not write any changes to your files until you have reviewed and approved them, ensuring you maintain full control over your codebase.
What kind of diff does Atlas show me?: Atlas presents a unified diff for every file edit. This standard format clearly highlights additions, deletions, and modifications line by line, making it easy to understand the exact changes Atlas intends to make.
Can I reject Atlas's proposed changes?: Yes, you have full control. When Atlas surfaces a unified diff for approval, you can choose to reject the proposed changes. This prevents Atlas from writing the file and allows you to guide the agent towards a different solution or take over the edit yourself.
Where in Atlas's code is this diff review implemented?: The core logic for computing and surfacing diff-reviewed edits is found in the Atlas-Backend. Specifically, it resides in the `atlas/src/tool/edit.ts` file, managed by the `EditTool` symbol.
Does Atlas review changes to new files?: Atlas computes a unified diff for every file edit, which includes the creation of new files (represented as an addition of all lines). This ensures that even new files proposed by Atlas are subject to your review and approval before being written to your project.
Is this feature available in the free core of Atlas?: Yes, Diff-reviewed edits are a fundamental safety and control mechanism within Atlas. As a proprietary tool with a FREE CORE, this capability is available to all users, ensuring a secure and controlled AI coding experience from the start.
How does this integrate with my existing git workflow?: Atlas reads git branches, status, and diffs, and can stage and create commits on your behalf. By requiring approval for every file edit, Diff-reviewed edits ensure that only approved changes are introduced, which you can then review again with standard git tools before committing, or let Atlas stage and commit for you.

Try Atlas in your terminal

The terminal-native AI coding agent. Free core, single binary.

Install Atlas

Related guides

Plan a Multi-File Change Before Editing with Atlas in 2026

How to plan a multi-file change with Atlas in 2026: the plan agent denies all edit tools, you research with codebase_search and lsp, then plan_exit hands off.

Atlas with Llama 4 Maverick: 1M Context on Open Weights in 2026

Llama 4 Maverick gives Atlas a 1M token context on open weights at $0.24 / $0.97 per Mtok on Bedrock or $0.20 / $0.80 on DeepInfra. Setup, tradeoffs, and fit.

Atlas for C in 2026

Atlas is a terminal-native AI coding agent for C in 2026. Run it in a project with a Makefile, have it find memory leaks or add Unity tests, and review the diff.

Atlas with StarCoder2 15B (Ollama): The Provenance Choice in 2026

StarCoder2 15B (Ollama) is BigCode's 9.1GB code model with a transparent training corpus and a 16K context, Free (self-hosted). Atlas setup and tradeoffs for 2026.

Atlas with Gemini 2.0 Flash-Lite: The Cheapest Google Model in the Registry (2026)

Gemini 2.0 Flash-Lite in Atlas: $0.075 per Mtok input, $0.3 per Mtok output, a 1,048,576 token context, an 8,192 token output cap, and no reasoning mode.

Atlas with Mistral Medium 3 (2505) in 2026: Symmetric Limits, $0.40 In

Mistral Medium 3 (2505) runs Atlas with symmetric 131,072 token context and output at $0.40 / 1M input tokens and $2.00 / 1M output tokens. Setup, limits, successors.

Atlas for Spring in 2026

Atlas, the terminal native AI coding agent, empowers Spring developers in 2026 with intelligent code assistance, secure local embeddings, and transparent review processes for enhanced productivity.

Atlas with Kimi K2.7 Code: Open Weights at Trillion-Parameter Scale (2026)

Kimi K2.7 Code drives Atlas at $0.95 / $4 per Mtok on a 262,144 token window. Open weights, 1T total parameters, served by four independent providers.

Browse this resource hub