Beta — 14-day free trial / 100,000 credits / no card required

Detect the error,
AI fixes it, 3-Retry guarantees it.

Receive errors from Sentry or your CI via webhook. Claude API analyzes the root cause and opens a fix PR automatically. The fix is iteratively improved up to 3 times until tests pass — all you have to do is review.

Red Codens is an automated code repair tool that uses AI to analyze production errors detected by Sentry or CI and automatically generates a fix pull request. Each fix is validated by running your test suite, and if a test fails, Red Codens automatically retries up to 3 times (3-Retry).

Get Started Free See How It Works
3-Retry
Auto-improvement loop on test failure
100 Files
Auto-collected from the dependency graph
24/7
Repair workers that never sleep

How It Works

From error detection to fix PR creation — the worker runs the whole flow end to end

1

Receive Errors via Webhook

Receive errors via webhook from any source — Sentry, GitHub Actions, CI, or your in-house monitoring tool. Stack traces and environment data are ingested as-is.

2

AI Root-Cause Analysis

Claude API (Sonnet) parses the stack trace using a systematic 6-step debugging methodology. Related files are auto-collected from the dependency graph, providing context across up to 100 files.

3

Generate Fix Code

Generate the smallest possible diff. The model understands your repository structure and architecture patterns, and follows the conventions of your existing codebase.

4

Run Tests & 3-Retry Improvement

Run tests; on failure, the test output is passed back as feedback to regenerate the fix. The AI auto-improves up to 3 times — only fixes that pass tests proceed to the PR step.

5

Auto-Create GitHub PR

The error summary, analysis results, and fix rationale are formatted into the PR body, then pushed to GitHub and a PR is opened. Reviewers just check the diff and merge.

Slack Notification

Notify Slack with the PR URL, analysis summary, test results, and confidence score. If you want to dig deeper, chat with the AI for follow-up questions.

The 3-Retry improvement loop that doesn't give up until tests pass

The first AI fix isn't always perfect.
Red Codens reads the test output, analyzes the failure systematically in 4 steps,
dynamically pulls in additional files as needed, and regenerates the fix.
Engineers no longer need to spend their valuable time on trial and error.

How the 3-Retry loop works
① Generate fix → Run tests
└ On failure: feed test output back, regenerate
② Pull in missing files dynamically → Re-test
└ Still failing → expand context for attempt 3
③ Up to 3 automated improvements
Only test-passing fixes become PRs

Key Features

Everything you need for auto-repair, from error detection to fix PR

Automatic Error Detection

Receive errors via webhook from any source — Sentry, GitHub Actions, CI, or your in-house monitoring tool. Detection-to-repair workflow runs automatically.

AI Root-Cause Analysis

Claude API (Sonnet) drills into stack traces with a systematic 6-step debugging methodology (analyze → hypothesize → trace → verify) to identify the true cause.

Dependency Graph Analysis

Build a dependency graph from import relationships and auto-collect related code across up to 100 files. Fixes account for repository structure and architecture patterns.

3-Retry Improvement Loop

On test failure, test output is passed back as feedback and the fix is auto-improved up to 3 times. On retry, missing files are dynamically added to context.

Automated Test Execution

Validate the fix with your project's test command before opening a PR. Only passing fixes reach review, dramatically reducing noise.

Automated GitHub PR Creation

The error summary, analysis, and fix approach are formatted into the PR body. The GitHub App pushes safely and opens the PR — reviewers just merge.

Bug Report Chat

Chat with the AI to dig deeper into analysis results. Ask "why was this fixed?" or "are there other affected areas?" and get answers grounded in full context.

External Report Portal

Route bug reports from external users or support teams through the same repair pipeline via the public bug report portal.

Codens Family Integrations

Pipe bugs found by Blue Codens (QA) into Red, then run auto-repair and re-test verification end to end.

How Red Codens Compares

An honest comparison with GitHub Copilot Autofix, Devin, and SonarQube

Based on publicly available information as of July 2026. Vendor specifications may change.

Dimension Red Codens GitHub Copilot Autofix Devin (Cognition) SonarQube + CI
Trigger Fires autonomously from production monitoring (Sentry/CI) — no instructions needed Only when CodeQL code scanning detects a vulnerability Started by a human assigning a task or ticket When static analysis runs in the CI pipeline
Scope of bugs Any production error (exceptions, logic bugs, etc.) Only known vulnerability patterns detected by CodeQL Whatever is assigned (not limited to bug fixes) Code quality / vulnerability findings matching static rules
Fix PR generation Fully unattended, all the way to an open fix PR Suggests fixes for vulnerabilities (CodeQL findings only) Opens PRs as an assigned task (not monitoring-driven) Flags issues only — does not generate fix code
Verification loop Auto-retries up to 3 times on test failure (3-Retry) Single-suggestion based; no public info on iterative retries Iterates as an agent, but outside a monitoring loop Not applicable (detection only)
Japanese support Native Japanese UI, analysis, and PR descriptions Multilingual, but not Japan-market specific Limited Japanese / Japan-market support Rule descriptions are multilingual

GitHub Copilot Autofix

Strengths: Native to GitHub with no extra install, and backed by CodeQL's mature vulnerability-detection database.

Copilot Autofix is scoped to security vulnerabilities detected by CodeQL — general logic bugs and exceptions are out of scope. Red Codens autonomously handles the full range of production errors detected via Sentry / CI.

Devin (Cognition)

Strengths: A general-purpose autonomous coding agent that can handle a wide range of tasks beyond bug fixing.

Devin operates on tasks or tickets assigned by a human; autonomous triggering from production monitoring is out of scope. Red Codens isn't "instruction-driven" — error detection itself is the trigger.

SonarQube + CI

Strengths: A proven static-analysis track record with broad language support and a low barrier to entry via the OSS edition.

SonarQube flags code-quality and vulnerability issues but does not generate fix code. Red Codens analyzes the root cause with AI and automatically opens a fix PR that has been validated by your tests.

Codens has a track record from its own dogfooding: 65% of our merged PRs are AI-generated (measured values published at www.codens.ai/stats). Red Codens processes each incident in roughly 3 minutes with an internal success rate of about 92%.

The Codens Suite

Codens is an AI development suite from Japan that automates the entire software development lifecycle — requirements (Green), implementation (Purple), code review & security audit (Orange), QA/E2E testing (Blue), and production error auto-fix (Red). Teams can build and operate software products without a large team of specialist engineers.

Green
PRD authoring
Purple
Implementation orchestration
Orange
Code review & security
Blue
QA automation
Red
Error auto-fix (this site)
Auth
Unified authentication

Track record: see Codens' development stats

Simple 4-Step Setup

Integrates with your existing Sentry, GitHub, and Slack right away

1

Create Account

Sign up at auth.codens.ai. Create an organization and invite teammates.

2

Configure Repos & Test Commands

Install the GitHub App and register target repositories along with a test command (e.g. `pytest`).

3

Distribute the Webhook URL

Configure the Red Codens webhook URL in Sentry, your CI, or monitoring tools. Slack integration can be enabled at this stage as well.

✓

Review the Fix PRs

When errors occur, just review and merge the auto-generated PRs. Recurring errors are tracked in the history.

Repair Run Example

Worker Progress

receive_error_webhook Done (0.4s)
analyze_root_cause Done (12.3s)
generate_fix (attempt 1) Done (8.1s)
run_tests Running…
open_pr Pending

AI Analysis Live Output

▸ Received Sentry webhook · KeyError: 'user_id'
▸ Building dependency graph (87 files)
▸ Loading repo structure (Clean Architecture / FastAPI)
› Hypothesis 1: auth middleware accessed before user resolution
› Tracing flow: routes/auth.py → middleware/auth.py
✓ Root cause: missing None guard in get_current_user
▸ Generating minimal fix…
▸ Running pytest tests/unit/test_auth.py
3-Retry auto-improvement PR only after tests pass Full audit log

UI Preview

See errors, fix proposals, and PR status in a single view

app.red.codens.ai/projects/example/proposals

Fix Proposals

Pending 3 Applied 12
FP-128 · Sentry · 2 min ago
fix: KeyError 'user_id' in auth middleware
3-Retry: 1/3 success · pytest 23 passed · confidence 92%
Awaiting Review
FP-127 · CI Failure · 18 min ago
fix: TypeError on /api/v1/orders date parsing
3-Retry: 2/3 success · pytest 87 passed · PR #421
PR Created
FP-126 · Sentry · 1 hour ago
fix: race condition in webhook deduplication
Analyzing: building dependency graph (87 files)
Analyzing
FP-125 · Report Portal · Yesterday
fix: locale fallback for invoice export
Merged · PR #418
Merged

Subscription Pricing

Start with a 14-day free trial, then choose from four plans with monthly credit allowances.

Credits are shared across the entire Codens family.

What are credits?

Red Codens consumes credits for AI error analysis, fix code generation, and test feedback improvement.
A monthly subscription gives you predictable operations, with flexible top-up packs available whenever you need extra credits.

Per-incident consumption depends on error complexity, repository size, and how many 3-Retry attempts are needed.

14-day free trial · 100,000 credits · no card required

Free Trial

For trying it out

¥0
100,000
credits (14 days)
No card required
Start free

Hobby

For solo developers

¥3,000 /mo
300,000
credits / month
5 seats
Get started
Popular

Pro

For teams

¥10,000 /mo
1,000,000
credits / month
10 seats
Get started

Business

For production workloads

¥50,000 /mo
5,000,000
credits / month
25 seats
Get started

Included in all plans

3-Retry improvement loop Dependency graph analysis Automated GitHub PRs Shared Codens family credits Audit log

For Enterprise (unlimited credits, custom contracts, SLA), reach out to support@corevice.com.

Free your engineers from late-night error response

Hand off everything from error detection to fix PRs to Red Codens. Your team can focus on review and design.

Get Started Free Contact Us

14-day free trial • 100,000 credits • no card required

Contact

Questions, enterprise plan inquiries, onboarding help — feel free to reach out

Email

We typically respond within 2–3 business days.

support@corevice.com

Other Codens Products

Consider PRD, QA, and development orchestration as a bundle.

Green Blue Purple
Start 14-day free trial (100,000 credits)

FAQ

Common questions before adopting Red Codens

What is Red Codens?

Red Codens is an AI automatic code repair system. It receives production errors from Sentry or your CI via webhook, analyzes the root cause with Claude API, generates fix code, validates it with tests, automatically opens a fix PR on GitHub, and notifies Slack — fully automated end to end. Each incident is processed in roughly 3 minutes, with an internal success rate of about 92%.

How does it fix bugs automatically?

Claude API (Sonnet) parses the stack trace with a systematic 6-step debugging methodology and auto-collects up to 100 related files from the dependency graph to pinpoint the root cause. It generates the smallest possible diff and runs your tests. On failure, the test output is fed back as feedback and the fix is auto-improved up to 3 times (3-Retry). Only fixes that pass tests are raised as GitHub PRs.

Which languages and error monitoring services (e.g. Sentry) are supported?

Any tool that can send a webhook works. Ingest error data from Sentry, GitHub Actions, CI errors, your in-house monitoring tool, or any other source. Because fixes are validated with your repository's test command (e.g. pytest), Red Codens supports a broad range of languages and stacks that can run tests.

How much does it cost (credits)?

A 5-tier subscription model (Free 14-day trial / Hobby ¥3,000 / Pro ¥10,000 / Business ¥50,000 / Enterprise), where each plan grants a monthly credit allowance. The entire Codens family (Red / Blue / Green / Purple) shares a unified organization credit pool. New signups start with a 14-day free trial (100,000 credits, no card required), and usage above the monthly allowance can be topped up via credit packs.

How does Red Codens handle security and permissions?

Writes to your repository are limited to safe, least-privilege pushes via a GitHub App, and every fix is raised as a PR for review before merge. Red Codens supports per-organization tenant isolation and permission management, full audit logs of all actions, and SSO via auth.codens.ai.

Do you support error sources other than Sentry?

Yes. Red Codens integrates with any tool that can send a webhook. Beyond Sentry, it also ingests GitHub Actions and CI failures, your in-house monitoring tools, and reports from the public bug report portal — all through the same repair pipeline. If you're on call and multiple monitoring systems fire at once, Red Codens absorbs all of them.

Are generated fix PRs merged automatically?

No — human review is the default. Red Codens only opens a GitHub PR once the fix passes tests, but it never auto-merges. The standard workflow has a reviewer check the diff and merge it themselves.