AI code review from the command line. Runs locally via Ollama or with any cloud LLM.

$ crev review --staged
using qwen2.5-coder:14b via Ollama (local)
reviewing 3 files (47 added, 12 removed)
context: Rich (4 types, 6 called fns, 2 tests)

[!] HIGH src/payments/processor.rs:142
    balance + amount can exceed i64::MAX when processing large transfers;
    use checked_add() and return Err on overflow

[~] MED  src/payments/processor.rs:98
    db.execute() result is silently ignored; if the INSERT fails, the caller
    receives a success response while the data was never written

[✓] LGTM src/utils/format.rs — change looks correct

3 findings (1 high, 1 med, 0 low) · 4.2s · qwen2.5-coder:14b

---

Requirements

• A git repository
• One of:
  • Ollama running locally (ollama serve + ollama pull qwen2.5-coder:7b)
  • ANTHROPIC_API_KEY, OPENAI_API_KEY, or GEMINI_API_KEY set

---

Install

curl -fsSL https://raw.githubusercontent.com/starc007/crev/main/install.sh | sh

Or build from source:

cargo install --path .

---

Quick start

# Stage some changes
git add -p

# Review them
crev review --staged

# Use a cloud model instead
crev review --staged --model claude-sonnet-4-6

# Install git hooks so every commit is reviewed automatically
crev init

---

Commands

crev review

Reviews code changes.

Options:
  --staged            Review staged changes (default)
  --unstaged          Review unstaged changes
  --commit <HASH>     Review a specific commit
  --commits <RANGE>   Review a commit range  e.g. HEAD~3..HEAD
  --model, -m <MODEL> Model to use (overrides config)
  --json              Output findings as JSON
  --fail-on <SEV>     Exit 1 if any finding at or above this severity [low|med|high]
  --security          Security-focused review mode
  --no-cloud          Never use a cloud LLM
  --path <PATH>       Path to git repo (default: current directory)

Examples

crev review --staged
crev review --commit abc1234
crev review --commits main..feature-branch
crev review --staged --model gpt-4o
crev review --staged --model claude-sonnet-4-6
crev review --staged --fail-on=high    # blocks git hooks on HIGH findings
crev review --staged --json            # machine-readable output
crev review --staged --security        # security-only scan

crev init

Installs git hooks and creates a .reviewrc config in the repo root.

crev init                        # install hooks + create .reviewrc
crev init --dry-run              # preview what would be installed
crev init --force                # overwrite existing hooks
crev init --hooks-only           # skip .reviewrc creation
crev init --uninstall            # remove hooks
crev init --ci                   # write .github/workflows/crev.yml
crev init --ci --model gpt-4o   # same, with OpenAI model + correct secret name

Installs two hooks:

• pre-commit: reviews staged changes before every commit
• pre-push: reviews all unpushed commits only when pushing more than one — single commits are already covered by pre-commit

crev history

Shows review history stored in a local SQLite database.

crev history              # last 10 reviews
crev history --patterns   # recurring findings (3+ times in 30 days)
crev history --clear      # delete history for this repo

History is stored at ~/Library/Application Support/crev/history.db on macOS and ~/.local/share/crev/history.db on Linux.

crev config

crev config --show    # print effective config for current directory
crev config --init    # create ~/.config/crev/config.toml with defaults

crev update

crev update    # upgrade to the latest release

---

Models and backends

crev supports four backends. The backend is auto-selected based on what's available.

Auto-detection order

1. Ollama running locally → use Ollama
2. ANTHROPIC_API_KEY set → use Anthropic
3. OPENAI_API_KEY set → use OpenAI
4. GEMINI_API_KEY / GOOGLE_API_KEY set → use Gemini

Selecting a model

The --model flag infers the backend from the model name:

# Ollama (local)
crev review --model qwen2.5-coder:14b
crev review --model deepseek-coder-v2:16b

# Anthropic
crev review --model claude-sonnet-4-6
crev review --model claude-opus-4-6

# OpenAI
crev review --model gpt-4o
crev review --model o3

# Gemini
crev review --model gemini-1.5-pro

API keys

export ANTHROPIC_API_KEY=sk-ant-...
export OPENAI_API_KEY=sk-...
export GEMINI_API_KEY=...        # or GOOGLE_API_KEY

OpenAI-compatible endpoints

Any OpenAI-compatible API (Groq, Together, local vLLM, etc.) works via OPENAI_BASE_URL:

export OPENAI_API_KEY=your-key
export OPENAI_BASE_URL=https://api.groq.com
crev review --model llama-3.1-70b-versatile

Ollama local models

Any model available in your Ollama instance works — crev picks the best one it finds automatically. Pull any coding model and crev will use it:

ollama pull qwen2.5-coder:14b
crev review --staged

Or specify explicitly:

crev review --staged --model qwen2.5-coder:14b

Set OLLAMA_HOST to point at a remote Ollama instance.

---

Configuration

crev init creates .reviewrc in your repo root. Commit it to share settings with your team.

[review]
# model = "qwen2.5-coder:14b"   # override auto-detected model
# backend = "anthropic"          # ollama | anthropic | openai | gemini
# api_key_env = "MY_API_KEY"     # env var to read the API key from (default per-backend)
max_tokens = 32000
severity_threshold = "low"      # low | med | high

[privacy]
strip_comments = false           # remove comments before sending to LLM
strip_strings = false            # replace string literals with <REDACTED>

[ignore]
paths = [
  "migrations/",
  "*.generated.rs",
  "vendor/",
  "*.pb.go",
]

[[rules]]
name = "no-raw-sql"
description = "All DB queries must use QueryBuilder, never raw SQL strings"

[[rules]]
name = "no-unwrap"
description = "Never use .unwrap() in non-test code — use ? or expect()"

Personal defaults (model choice, API keys) go in ~/.config/crev/config.toml — this file is never committed.

Config lookup order: .reviewrc (current dir → upward) → ~/.config/crev/config.toml → built-in defaults.

Output formats

Terminal (default)

A spinner shows while the model analyzes. Each finding streams to the terminal as soon as its line is complete:

⠴ analyzing...

[!] HIGH src/payments/processor.rs:142
    balance + amount can exceed i64::MAX — use checked_add()

[~] MED  src/auth/session.rs:67
    session token written to log at info level — strip before logging

[i] LOW  src/utils/retry.rs:23
    fixed sleep of 1000ms; use exponential backoff to avoid thundering herd

3 findings (1 high, 1 med, 1 low) · 4.2s · qwen2.5-coder:14b

JSON (--json)

{
  "findings": [
    {
      "severity": "High",
      "file": "src/payments/processor.rs",
      "line": 142,
      "message": "balance + amount can exceed i64::MAX — use checked_add()"
    }
  ],
  "github_annotations": [
    {
      "path": "src/payments/processor.rs",
      "start_line": 142,
      "end_line": 142,
      "annotation_level": "failure",
      "message": "balance + amount can exceed i64::MAX — use checked_add()"
    }
  ]
}

github_annotations is compatible with the GitHub Check Runs API and appears as inline comments on PR diffs.

---

CI / GitHub Actions

Write a ready-to-use workflow with one command:

crev init --ci                   # uses Claude (ANTHROPIC_API_KEY)
crev init --ci --model gpt-4o   # uses OpenAI (OPENAI_API_KEY)
crev init --ci --model gemini-2.0-flash  # uses Gemini (GEMINI_API_KEY)

This writes .github/workflows/crev.yml directly and prints the exact secret name to add:

Created .github/workflows/crev.yml

Add your API key as a repository secret:
  Repo → Settings → Secrets and variables → Actions → New repository secret
  Name: ANTHROPIC_API_KEY

Then commit and push:
  git add .github/workflows/crev.yml
  git commit -m "ci: add crev code review"

What happens on a PR:

1. PR opened → crev reviews the diff automatically
2. Findings posted as inline comments on the PR diff
3. A summary comment posted on the PR thread (updated in-place on re-runs — no spam)
4. Merge blocked if any HIGH findings exist

Re-running manually: comment /crev on any PR — the workflow triggers and updates the existing comment with fresh results.

Triggers:

Event	Behaviour
PR opened	Runs automatically
/crev comment	Runs on demand, reacts with 👀 to acknowledge
Push to PR branch	Does not run — comment /crev to re-review

---

Pattern detection

crev tracks every review in a local SQLite database. After the same finding appears 3+ times in a 30-day window, it warns you before the next review:

⚠ Recurring pattern detected (5x in 30 days):
  integer overflow in arithmetic operations

This surfaces systemic issues that keep slipping through code review.

---

License

HEHEHEHEHE