Skip to content
ATAI Today Brief
HomeNewsConceptsGuidesToolbox
AboutSubscribeUA
Subscribe

AI Today Brief

The daily AI-engineering brief. Built in public. EN · UA.

XTelegramLinkedInYouTubeRSS
NewsConceptsGuidesSubscribeAdvertiseAboutEditorial policyAI disclosurePrivacyTerms

© 2026 AI Today Brief. All rights reserved.

  1. Home/
  2. News/
  3. Agents & MCP/
  4. Review-flow: Automate 80% of code reviews using Claude Code and Model Context Protocol
Agents & MCP

Review-flow: Automate 80% of code reviews using Claude Code and Model Context Protocol

July 4, 2026· 5 min read
OKCurated by Oleksandr Kuzmenko, AI Product Engineer·Updated July 4, 2026·Sources cited on every story
AI-assisted · editor-reviewed·How we use AI
Agents & MCP

Review-flow is an open-source server that automates code review pipelines on GitHub and GitLab. It utilizes Claude Code background sessions and a dedicated Model Context Protocol server to execute structured multi-agent audits.

Impact: High

Why it matters

Self-host Review-flow to automate repetitive code quality, architecture, and security reviews directly within your merge requests.

TL;DR

  • 01Runs multi-agent reviews (SOLID, Testing, DDD) asynchronously on GitHub and GitLab.
  • 02Uses isolated Git worktrees and a memory guard capped at 4 GB RSS for stability.
  • 03Leverages typed MCP commands instead of text parsing to manage Git thread state safely.

Key facts

4 GBMax RSS Memory
15 minutesHard Timeout
Max RSS Memory
4 GB
Hard Timeout
15 minutes
Default Concurrency Limit
2 parallel reviews

Background Process Orchestration

Review-flow replaces interactive CLI loops with detached background sessions. When a webhook triggers, the server spawns a background worker via claude --bg and immediately returns. Completion of the review is detected via a three-way race condition with first-wins semantics: 1. The MCP skill calls set_phase('completed') on the local server. 2. Polling claude agents --json every 30 seconds shows completion. 3. A hard backstop timeout of 15 minutes is reached.

Once completed, the report is read from the isolated worktree directory (.claude/reviews/report-<mrNumber>.md) and posted to the Git platform. The session is closed using claude stop && claude rm.

Self-Hosted Infrastructure Specifications

Review-flow enforces system safety and cost controls locally:

  • Memory Guard: Automatically terminates running review sessions if the RSS memory consumption exceeds 4 GB.
  • Concurrency Control: Limits parallel reviews (default is capped at 2 parallel tasks).
  • Billing Limits: Tracks monthly token usage from transcripts to block execution if budgets are exceeded.
  • Worktree Lifecycle: Runs daily sweeps to prune worktrees of closed MRs older than 24 hours or with modification times exceeding 7 days.

Try it in 2 minutes

reviewflow init --yes

bash

✓ When to use

  • When managing high-velocity teams looking to offload basic architectural and quality checks directly inside merge requests.

✕ When NOT to use

  • When security policies strictly forbid running self-hosted web servers connected to third-party LLM providers.
  • When your team lacks a dedicated server environment capable of running background Node.js processes.

What to do today

  • →Initialize Review-flow locally by running `reviewflow init --yes` in your terminal.
  • →Configure webhooks on your GitLab/GitHub projects pointing to your self-hosted Review-flow instance.
#Claude Code

Sources

  • Review-flow GitHub Repository
ShareShare on XShare on LinkedIn
← Previous storyAnthropic investigates potential macOS session and cache leakage in Claude CodeNext story →NVIDIA HORIZON uses git worktrees and prompt caching to build hardware agents

Related stories

  • Agents & MCPArkon: Self-Hosted Enterprise Knowledge Hub and Model Context Protocol Server
  • Agents & MCPApache Magpie Introduces Vendor-Neutral Agent Recipes for Repository Maintainers
  • Agents & MCPSimon Willison Launches llm-coding-agent Python Library via Claude Code Spec-Driven TDD
  • Agents & MCPDeploy and debug Model Context Protocol servers in production with Manufact Cloud

Email digest

Get the morning AI brief

One email a day — the stories that matter for engineers, founders and tech leads. Human-edited, with links to primary sources.

  • ✓120+ sources scanned daily
  • ✓Edited by a human
  • ✓1 email per day
  • ✓EN + UA

By subscribing you agree to the privacy policy.