I Built AgentCheck Because “The Coding Agent Said Done” Wasn’t Enough

A developer built AgentCheck, a deterministic tool that verifies coding agents' changes by comparing the Git repository state before and after the agent runs, without relying on an LLM for analysis. The tool creates a checkpoint using a temporary Git index, then reports changes, findings, risk score, and a verdict to guide human review before committing.

AI coding agents are getting surprisingly good at writing code. I use them regularly, and they can handle increasingly large tasks: refactoring code, adding features, updating dependencies, modifying configuration, creating migrations, and touching files across an entire repository. But I kept running into the same problem after the agent finished: How do I independently verify what it actually changed? The agent usually gives me a perfectly reasonable summary. Something like: Done. Implemented the requested changes, updated the tests, and cleaned up the affected code. Useful? Absolutely. Enough for me to commit without checking? Not really. So I built AgentCheck . After a coding agent finishes a task, I still find myself manually checking things like: Of course, Git already gives us the raw information. I can run: git status git diff git diff --stat Then inspect individual files. And I still do that. But once coding agents become part of your normal workflow, repeating the same verification process after every task starts to feel like something that should be structured. That was the idea behind AgentCheck. AgentCheck creates a trusted checkpoint before your coding agent starts working. Then, after the agent finishes, it compares the current Git-visible repository state with that checkpoint. The basic workflow is deliberately small: agentcheck start Then let your coding agent work. That can be: When the work is finished: agentcheck AgentCheck then produces four sections: Changes Findings Risk Verdict For example: AgentCheck Changes ──────────────────────────── 2 modified 1 created 0 deleted 0 renamed A packages/example/new-file.ts M package.json M src/example.ts Findings ──────────────────────────── ⚠ Dependency change detected. Risk ──────────────────────────── Score: 3 — MEDIUM Verdict ──────────────────────────── REVIEW RECOMMENDED The important part is that this result comes from the repository state itself — not from the coding agent's explanation of what it believes it changed. This was one of the main design decisions. There are already many AI code-review tools, and some of them are very capable. But that wasn't the problem I wanted AgentCheck to solve. If one LLM changes my repository, I didn't necessarily want the verification layer to be: LLM changes code ↓ another LLM reviews the first LLM I wanted a smaller and more predictable layer: Coding agent ↓ Actual Git-visible changes ↓ Deterministic checks ↓ Human review ↓ Commit So AgentCheck does not use an LLM for its analysis. The checks are deterministic. Given the same repository state, AgentCheck should produce the same result. The first public version intentionally keeps the scope limited. AgentCheck can currently highlight things such as: These signals feed into a transparent risk score and a restrained verdict. For example: 0–2 → LOW 3–6 → MEDIUM 7+ → HIGH The goal is not to say: This code is correct. AgentCheck cannot know that. The goal is closer to: These are the parts of this change set that probably deserve your attention before you commit. One technical requirement was particularly important to me: AgentCheck should not modify the developer's actual Git index, working tree, or history. The checkpoint implementation uses Git's tree/index model with a temporary alternate index. Conceptually: Current repository state ↓ temporary Git index ↓ git write-tree ↓ checkpoint tree Later, AgentCheck creates another representation of the current state and compares: checkpoint tree ↓ diff ↑ current tree This means the developer can already have: when the checkpoint is created. Those pre-existing changes become part of the baseline rather than being incorrectly attributed to the coding agent. The real Git index remains untouched. AgentCheck currently has: The verification happens locally. That also keeps the workflow simple: npm install -g @agentcheck/cli agentcheck start coding agent works agentcheck There is also a VS Code extension if you prefer reviewing the result inside the editor. I developed AgentCheck primarily using Codex , but I intentionally avoided coupling AgentCheck to any specific coding-agent product. It doesn't need to understand the agent session. It doesn't need an agent plugin. It doesn't need the agent to tell AgentCheck when it is finished. AgentCheck only cares about the resulting repository changes. So the same workflow can sit after: Claude Code Codex Cursor another coding agent That separation is important to me. Coding agents will change. The Git repository remains the source of truth. AgentCheck is currently available as both a CLI and a VS Code extension. Install: npm install -g @agentcheck/cli Then: agentcheck start and later: agentcheck The extension exposes the same review model inside VS Code: CHANGES FINDINGS RISK VERDICT The VS Code extension is intentionally a thin UI over the same deterministic core rather than a separate analysis engine. AgentCheck is open source under the Apache License 2.0 . GitHub: https://github.com/emreordu/agentcheck https://github.com/emreordu/agentcheck npm CLI: https://www.npmjs.com/package/@agentcheck/cli https://www.npmjs.com/package/@agentcheck/cli npm Core: https://www.npmjs.com/package/@agentcheck/core https://www.npmjs.com/package/@agentcheck/core VS Code Marketplace: https://marketplace.visualstudio.com/items?itemName=agentcheck.agentcheck-vscode https://marketplace.visualstudio.com/items?itemName=agentcheck.agentcheck-vscode The first release was mainly about proving the checkpoint and deterministic verification model. For the next version, I'm currently exploring things such as: One thing I'm deliberately trying to avoid is turning AgentCheck into a giant AI code-review platform. I want the core idea to remain simple: Independent verification of what actually changed. AgentCheck is still early. The most useful feedback for me right now isn't: Add more features. It's things like: If you use Claude Code, Codex, Cursor, or another coding agent in real repositories, I'd love to hear how this approach fits into your workflow. Try it. Break it. Tell me what it gets wrong. Don’t trust “done”. Verify the result.