LiveReview: Blast-Radius Aware AI Code Review for Business-Critical Systems

LiveReview is an AI code reviewer that scores every hunk of a diff by blast radius: how far a change reaches through your call graph, how much persistent state it touches, and how well-tested it is. A 3-line change to a shared auth check can outrank a 300-line UI tweak. Your team's attention goes to the highest-risk code first, not spread evenly across every diff.

LiveReview's Blast Radius & Review Priority scoring, live in the diff viewer.

The exact math, not a black box	Visualize blast radius at a glance	Every factor that feeds the score

Self-Host for Free: Get Started in 5 Minutes
Want a guided rollout? Join the 14-Day Transformation Program

🚀 Get Started

📊 See the Product In Action

🔌 Integrate LiveReview

💰 Pricing & Trust

🤝 Get Hands-On Help

Get LiveReview running in under 5 minutes. Self-hosting starts free, with the same 30k LOC/month, bring-your-own-key plan available on the cloud. You can scale up from there. See Pricing & Enterprise for the full breakdown.

# Quick demo setup (localhost only, no webhooks)
curl -fsSL https://raw.githubusercontent.com/HexmosTech/LiveReview/main/lrops.sh | sudo bash -s -- setup-demo

Requirements:

• bash shell (zsh/fish not officially supported)
• Ubuntu/MacOS
• Docker (with docker-compose or docker compose)
• jq

You'll need a Free Licence to get started. Follow the guide here.

Access your installation:

• Web UI: http://localhost:8081/
• API: http://localhost:8888/api

For teams needing external access and webhooks, follow the Productionization Guide.

AI writes code faster than any human can review it by hand. LiveReview gives you a checkpoint at every stage where that code could reach production. Turn on the stages that matter to you, per repository.

Not every repo needs the same amount of scrutiny. Turn on more checkpoints where the blast radius of a bad change is high, and fewer where speed matters most.

• Commit Only: fastest, lightest net. Fine for low-stakes, throwaway repos.
• Commit + Scheduled: the startup pick. Near-zero friction day-to-day, plus a daily sweep that catches anything that slipped past commit-time checks.
• Commit + Before Push + MR/PR: the standard team flow. A human sees every change before it merges.
• All Five (Full Harness): Commit, Before Push, MR/PR, CI/CD, and Scheduled together, for repos where a bad change is expensive: payments, auth, core infra.

There's no single right answer, only the right trade-off for a given repository, team, or organization. Mix and match per repository, and change your mind any time.

Setting up MR/PR reviews for your provider? See the step-by-step guides for GitHub, GitLab, Bitbucket, Gitea, and Azure DevOps.

Every commit git-lrc reviews gets checked against the same risk categories LiveReview tracks across production codebases:

See it live: Analyze Findings with the Impact Report and Export Impact Reports as PDF or CSV.

Correctness → Business Rule Violations A discount, limit, or policy nobody approved gets applied automatically, at scale.

Security → Authentication A weak login flow is an open door, and attackers check every door.

Maintainability → Code Complexity Code only one person understands is a single point of failure with a name and a vacation schedule.

Ask Livi a product, engineering, or ops question in plain English. Livi answers with a chart pulled straight from your organization's own data. No dashboards to build, no SQL to write.

Every engineering decision becomes more data-backed, so you can act with confidence instead of guesswork.

The same 7 categories also power a one-click Onboarding Report, exportable as HTML or PDF, with real charts pulled from your own review history:

Livi's answers reach you where you already work: watch Generate Engineering Reports via Slack and Get Engineering Reports in MS Teams. For the reasoning behind this, see Understand Engineering Decisions and Generate Engineering Reports in the MCP docs.

Below is a sample of the questions different roles ask Livi. Each chart uses the same specs as the interactive demo on the live site.

Persona	Question to Livi	Chart
Exec	"Is LiveReview adoption increasing across the org?"
Exec	"Which repositories have adopted LiveReview the most?"

Persona	Question to Livi	Chart
Product	"Which repos are gaining or losing engineering velocity?"
Product	"How much code are we reviewing in each repository?"

Show 12 more examples (Engineer Analysis, Review Quality, Cost & Efficiency, Engagement & Trust, Summary & Comparison, Trace & Investigate)

Engineer Analysis

Persona	Question to Livi	Chart
Eng Manager	"Who are our top contributors by review volume?"
Eng Manager	"How does each engineer trigger their reviews?"

Review Quality

Persona	Question to Livi	Chart
Eng Manager	"What are the most concerning issue types this quarter?"
Eng Manager	"Are engineers actually incorporating reviews into their daily workflow?"

Cost & Efficiency

Persona	Question to Livi	Chart
Exec	"How much does LiveReview cost us per day?"
Product	"Which AI provider gives us the best value?"

Engagement & Trust

Persona	Question to Livi	Chart
Eng Manager	"Are people trusting the reviews LiveReview produces?"
Product	"Which engineers get the most value from LiveReview?"

Summary & Comparison

Persona	Question to Livi	Chart
Exec	"How does this week compare to last week?"
Eng Manager	"What's the overall severity mix across all our findings?"

Trace & Investigate

Persona	Question to Livi	Chart
Engineer	"Show me reviews connected to last week's production incident."
Engineer	"Which files keep showing up with issues?"

Charts above use sample data, for illustration only. Try the fully interactive version, with live chart drill-down, at hexmos.com/livereview.

Most AI review tools flag style nits and treat every line the same. LiveReview is built around three things most tools skip:

• Blast-Radius scoring: every hunk is ranked by how much of the system it can actually break, so reviewers spend their limited time on the change that could take down production, not the one that renamed a variable.
• A named taxonomy of 104 failure patterns across Reliability, Correctness, Security, Compliance, Maintainability, and Cost (see Prevent Outages, Breaches, and Technical Debt), checked on every commit, not just at PR time.
• Livi, an AI chatbot for your engineering data: ask any question in plain English — adoption, cost, quality, who's actually incorporating review feedback — and get a data-backed chart back, not a guess. See Data-Backed Decisions.

Blast-Radius scoring and the failure taxonomy change what gets reviewed. Livi changes what gets decided: every rollout, staffing, or process call is backed by real numbers pulled from your own review history, not gut feel. That combination is what leads teams to keep it turned on:

• Accelerate Delivery Cycles: Cut PR review time from hours to minutes, because reviewers see what matters first instead of reading top to bottom.
• Save Senior Engineering Time: Free senior developers from routine reviews. Let them focus on mentorship and high-impact architecture work.
• Drive Quality Excellence: Track metrics that show improvements in code standards, fewer defects, and better development efficiency.
• Decide with Confidence, Not Guesswork: Ask Livi instead of guessing. Every engineering, staffing, or process decision gets a chart pulled from real review history behind it.

See where reviews get stuck (Sankey flow from open to merged) and where issues cluster by category (treemap), pulled from your own review history.

Review pipeline, at a glance	Issue distribution by category

LiveReview comes with its own fine-tuned AI model, ready from day one. Prefer your own provider? Bring your own key (BYOK) for Gemini, OpenAI, AWS Bedrock, a self-hosted Ollama model, or any other LLM. See the AI Integration guide, or watch Connect Google Gemini and Gemini Enterprise, Connect Amazon Bedrock, or Connect DeepSeek, OpenRouter, OpenAI, and Ollama.

Connect a repository from GitHub, GitLab, Bitbucket, Gitea, or Azure DevOps, and LiveReview reviews it the same way, with the same Blast Radius scoring. Watch Connect GitHub to LiveReview or Connect Self-Hosted GitLab, or see the full Git Provider setup guide.

Browse every repository and merge or pull request LiveReview can see, in one list, no matter which git provider it lives on. Trigger a review straight from that list — see Trigger PR Reviews from the Dashboard.

Every connected repository	Every merge/pull request

Show 7 more features (review list, progress tracking, custom prompts, team learnings, PR summaries, AI clarification, scheduled reviews)

View All AI Reviews in One Place

See every review's status, from queued to complete, and jump straight into the ones that need attention. See Trigger Manual Pull Request Reviews.

Track Which Files Are Reviewed, Live

Watch a review work through your diff file by file, so you know exactly what's covered and what's still queued.

Customize Review Prompts to Fit Your Team

Write custom prompts so the AI reviewer enforces your team's own coding standards and priorities, not generic defaults. (Premium & Enterprise) Watch Customize AI Review Prompts for Your Team, or read Customize LiveReview to Your Team's Best Practices.

Discuss with AI in MR and See it Learn Everyday

Every discussion in a merge request becomes a stored "learning": a best practice, a recurring issue, or a team convention the AI applies to every future review. See it in Improve Reviews with Organizational Learning.

Sharp AI-Generated Pull Request Summaries

Every pull request gets a summary of what changed, why it matters, and what risks were flagged, so reviewers don't have to read the whole diff to know where to look. See Ask Questions About Code via Inline PR Comments.

Ask AI for Clarification or Debate Code Changes

Reply to any AI comment in the merge request to ask why it flagged something, or push back on it. The AI has the full diff context, not just the one line it commented on. Watch Reply to AI Review Comments and Get Guidance, and Auto-Fix Review Issues with Claude Code or AI Agents.

Scheduled Reviews: A Safety Net for the Code Nobody Reviewed

Not every change goes through a full review:

• A hotfix might land straight on the main branch.
• A dependency bump might merge on its own.

For a small, fast-moving team, that's often the right call, you can't review every line by hand and still ship fast. Scheduled Reviews are the safety net for exactly that gap.

• Checks on its own schedule. LiveReview reviews your default branch even when nobody asked it to, and catches anything that got in outside your normal commit, push, or PR checks.
• Per-repository control. Turn it on with one toggle.
• Your own cadence. Pick how often it runs, in plain cron syntax, or leave it blank and LiveReview checks once a day.
• Always visible. See the last time it ran and the next time it will, right in the schedule list.
• Zero upkeep. Runs by itself in the background, nobody has to remember to trigger it.

For most teams, once a day on the main branch is enough to keep quality high without slowing anyone down.

The schedule list, per repository	Editing a repository's schedule

Install git-lrc for commit-time reviews in any terminal. Use claude-lrc when you build inside Claude Code. Both tools share the same AI review engine and the same monthly LOC quota.

# Typical Git Guardrails Flow
git add .
git lrc review

# Or skip explicitly (auditable in git log):
git lrc review --skip
git commit -m "message"

git-lrc (git commit hook):

• Git-native, works in any repo without a cloud platform connection
• Skips tracked in git log, auditable, not silent
• One-line install, 30k LOC free every month

claude-lrc (Claude Code integration):

• Review, vouch, and skip inside Claude Code without leaving the chat surface
• Natural language or slash commands (/lrc:review, /lrc:skip, /lrc:vouch)
• Bundled with git-lrc, no separate install needed

Full CLI docs: Getting Started · Reviewer Workflow · Repository Rules · Security

Linux/macOS:

curl -fsSL https://hexmos.com/lrc-install.sh | bash

Windows (PowerShell):

iwr -useb https://hexmos.com/lrc-install.ps1 | iex

Watch the One-Line Installer for LiveReview Self-Hosted demo, or follow the full install guide.

No local setup needed. Paste a merge or pull request URL into LiveReview and it runs the same review, with the same Blast Radius scoring.

git-lrc and claude-lrc are the two CLI tools above, in motion. Both plug into the same LiveReview backend.

git-lrc catching real issues on commit: leaked credentials, expensive cloud calls, and sensitive data in log statements, all flagged before the commit lands.

claude-lrc reviewing a diff inside Claude Code: no separate terminal, no context switch, just a slash command or a plain-English request.

Issue Navigator: every finding, filterable by severity, category, and subcategory, with a one-click send to your AI agent.

Summary Deck: a 60-second slide summary of what changed, why, and what risks were flagged, generated automatically for every review.

Risk-Scored View: every hunk ranked by blast radius and customer-impact potential, with the full signal breakdown behind each score.

Score badge, with breakdown	Whole diff, ranked by risk

Connector management: switch AI providers, or reorder them to set review priority, from one screen.

Get AI code reviews without leaving your editor. Available for VSCode, Cursor, and Antigravity.

Every code review LiveReview performs adds to a growing source of engineering intelligence. LiveReview exposes this two separate ways, for two separate purposes:

• MCP Server — for AI assistants and agents (Claude, Cursor, Windsurf) that need to ask questions and take action conversationally.
• REST API — for scripts, CI/CD pipelines, and custom integrations that need direct HTTP calls with no AI agent in the loop.

Both let you:

• Generate custom reports
• Identify your strongest contributors
• Uncover quality and security trends
• Drill into engineering activity in minutes, not hours

Real use cases from teams already doing this: Prevent Production Issues, Turn Findings into Tickets, Keep Project Management in Sync, and Generate Release Notes — see the full MCP use-case list.

The same key authenticates both the MCP server and the REST API.

1. Go to LiveReview
2. Click on Settings
3. Navigate to API Keys
4. Generate and copy a new API key

Watch Create and Manage API Keys.

For AI assistants and agents: Claude Desktop, Claude Code, Cursor, Windsurf, or anything else that speaks MCP.

Add the following block to your MCP client's configuration file:

• For eg: Claude Desktop: claude_desktop_config.json
• Other clients: Check the client's documentation for the equivalent file.

{
  "mcpServers": {
    "livereview": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://livereview.hexmos.com/api/mcp",
        "--header",
        "X-API-KEY: ${LIVEREVIEW_API_KEY}"
      ],
      "env": {  
        "LIVEREVIEW_API_KEY": "<YOUR_LIVEREVIEW_API_KEY>"  
      }                       
    }
  }
}

Replace <YOUR_LIVEREVIEW_API_KEY> with your actual LiveReview API key. See the MCP Configuration docs, or watch Connect LiveReview MCP Server to AI Coding Assistants.

Running an AI agent in your CI/CD pipeline instead of a chat assistant? The MCP server works there too: Automate Code Reviews in CI/CD with LiveReview MCP.

Once connected to the MCP server, you can ask your assistant to interact with LiveReview. Each MCP tool below wraps one REST API endpoint (its name follows the endpoint's method and path), but the tool itself is only reachable through the MCP server, not by calling the endpoint directly.

Full MCP reference: MCP Usage docs

For scripts, CI/CD pipelines, and custom integrations that call LiveReview directly over HTTP, with no AI agent or MCP client involved. Same API key as above, sent as the X-API-KEY header. Covers reviews, reports, learnings, billing, connectors, and more.

Full REST API reference: hexmos.com/livereview/docs/livereview/api

A good reviewer knows your language and framework. A great reviewer also knows your repository: which patterns your team prefers, which dependencies are off-limits, and which files don't need a second look. Drop a .lrc/ directory in your repo, and LiveReview reads it on every review.

.lrc/
├── ignore               # files the reviewer never sees
├── rules/
│   ├── INSTRUCTIONS.md  # read first, every review
│   ├── security.md
│   └── style.md
└── policy/
    └── tools.toml       # which checks are allowed to run

Full reference: Repository Rules docs · Set Review Rules

Adaptive Reviews uses two AI models instead of one: a powerful Leader Model finds complex issues, and a cost-efficient Helper Model explains them. Same review quality, at half the price.

See it explained: Adaptive Reviews: Cut AI Costs by 40-50%.

LiveReview's pricing tracks reviewed workload, not headcount:

LOC (Lines of Code) means the code shown in the reviewed diff, not your total repository size. Paid plans keep users unlimited, so cost tracks reviewed workload, not seats.

Custom deployments, SSO integration, dedicated AI keys, and priority SLA support, for scaling engineering organizations:

• Security & Ops: Self-hosted deployment (optional), support for multiple organizations, custom domain hosting, SSO & User Directory integration (SAML/OIDC), and full data privacy
• Flexible AI & Models: Connect to private cloud LLMs or opt for fully self-hosted AI models using Ollama or your private infrastructure to guarantee no code leaves your network
• Custom Integrations: Custom API access, bespoke workflow integrations, and engineering insights dashboards tailored to your development tooling
• Dedicated SLA Support: Prioritized support channel with dedicated SLAs, custom development, and professional onboarding

Get a Self-Hosted License · Explore Self-Hosted Enterprise · Apply a Licence

Watch Enterprise License Management in action.

The LiveReview Docs go far deeper than this README. A sample of what's there:

Self-Hosted Setup

• Wiki Overview · Download, Install and Run LiveReview · Productionize LiveReview · lrops.sh Reference
• Get a Licence · Apply a Licence · Backup LiveReview · Update LiveReview
• Create Your First Review · Add Your Team · Run a Secure Self-Hosted AI Code Review Powered by Ollama

Git & AI Provider Integration

• Adding Git Providers: GitHub · GitLab · Bitbucket · Gitea · Azure DevOps
• Add AI Integration: Google Gemini

Git-Native CLI (git-lrc / claude-lrc)

• Getting Started · Install · Concepts: Workflow, Roles, Collaboration
• Repository Rules · Set Review Rules · Integrations · git-lrc Security

MCP Server (for AI assistants and agents)

• MCP Configuration · MCP Usage
• Use cases: Prevent Production Issues · Turn Findings into Tickets · Keep Project Management in Sync · Generate Release Notes · Generate Engineering Reports · Understand Engineering Decisions — full list

REST API (for scripts, CI/CD, and custom integrations, no AI agent required)

• Full API Reference: reviews, reports, learnings, billing, connectors, and more

Video Library

• hexmos.com/livereview/demo has dozens of short, focused demos, filterable by role (Developer, Engineering Manager, CTO, CEO): setup, every git and AI provider integration, review workflows, reporting and Slack/Teams automation, and team administration.

FAQ & Security

• Docs FAQ · Full Security Documentation

LiveReview documents the questions enterprise teams ask first: deployment model differences, code and data handling, AI safeguards, supply-chain visibility, and security response timelines.

• Separate guidance for self-hosted/Ollama and cloud LLM deployments
• Explicit data handling: what leaves your network, when it happens, and retention/deletion expectations
• Prompt-injection and unsafe-output mitigations, automated scanners, SBOM visibility, and transparent GitHub source with responsive disclosure policy

Open Security Page · Read SECURITY.md · Report a Vulnerability

What data is collected, stored, and used in LiveReview?

• Only the diff is sent to the AI model. Nothing else.
• Your code is never stored, and never used to train any AI model.
• Own codebase is scanned continuously with Gitleaks, OSV Scanner, Govulncheck, and Semgrep, via GitHub Actions.
• A Bill of Materials (SBOM) is published with every release.

How does self-hosted deployment differ from cloud in terms of security?

• Self-hosted: your team runs the entire stack and database. Your infra team controls storage, backups, retention, and network access.
• Cloud / provider-integrated: LiveReview sends data to configured external provider endpoints for AI inference and git provider operations.

Does LiveReview provide a Software Bill of Materials (SBOM)? Yes, generated automatically on every release using Syft, published to GitHub release assets.

Is LiveReview SOC 2 Type II certified? Not at this time. Security docs, scan history, SBOM, and source code are all public instead, so enterprise buyers can review the actual posture rather than take a certification on faith.

For complete details, including local security scan commands and how to enable the gated-off scanning workflows, see SECURITY.md. For pricing, LOC, and general product questions, see the FAQ below.

What is Blast-Radius scoring, exactly? A per-hunk score combining:

• Call-graph reach and cross-package impact
• Persistent-state mutation
• Cyclomatic and cognitive complexity
• Test coverage gaps

Shows which parts of a diff can do the most damage if something is wrong, so reviewers spend their limited attention where it matters most.

What is LOC in LiveReview? Lines of Code shown in the reviewed diff — not your total repository size.

How much LOC is available on the premium plan?

All paid bands keep users unlimited and refresh every month.

What does the enterprise plan offer?

• Multiple organization support
• SSO & directory sync (SAML/OIDC)
• Self-hosted deployment within your own infrastructure
• A custom domain
• Full data privacy over where your code and review data are stored

Is my code secure with LiveReview? Only the diff is sent to the AI model, nothing else. Your code is never stored or used to train any model. See Security above for full details.

What's the difference between self-hosting this repo and the cloud version?

• This repo: the self-hosted product itself, Docker-based, runs entirely on your infrastructure.
• Cloud version: the same review engine as a managed service, so you don't run the stack yourself.

LiveReview is distributed under a modified variant of Sustainable Use License (SUL).

For detailed terms, examples of permitted and prohibited uses, and definitions, see the full LICENSE.md.

Self-hosting gets you the tool. The 14-Day Transformation Program gets your whole team, from execs to individual contributors, actually using it well. It covers onboarding, workflow integration, and measurable before/after engineering metrics.

Join the 14-Day Transformation Program →

Self-Host: Get Started Free | Cloud: Try hexmos.com/livereview | Documentation