# Codens — Full Content (plain text, for AI agents)

> Updated: 2026-07-16
> This file mirrors the full text content of www.codens.ai (home, pricing,
> comparison, and FAQ) in a single plain-text document for LLM consumption.
> For a short overview, see /llms.txt. For the help/agent reference site,
> see https://help.codens.ai/llms-full.txt

================================================================================
## HOME (https://www.codens.ai/)
================================================================================

### Canonical definition

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.

### Headline metrics

- Development lead time: -80% (shorter planning-to-release cycle)
- Bug recurrence rate: -90% (AI auto-fix x test automation)
- Engineering effort: -70% (PRD generation, review & analytics automated)
- Ship velocity: +5x (actual result: 35 tickets, est. 1-2 person-months,
  completed by Purple Codens in 16h45m on 2026-04-20; final numbers
  subject to CEO review)

### Dogfooding stat

65% of Corevice's own merged PRs are AI-generated (measured via the
GitHub API, updated weekly). Median PR-to-merge time: 2 minutes. 93% of
PRs merge within 1 hour. Live dashboard: https://www.codens.ai/stats/

### Case studies

Portament Inc.:
- 5 projects running in parallel
- Business members co-review requirements with Green/Red Codens and log
  tickets
- Purple Codens implements immediately — zero engineer wait time
- Requirements don't need to be finalized first; the team implements and
  iterates while running
- Engineers are escalated to only when needed, debugging via MCP directly
- Result: 3 business members + 1 engineer, 1.5 months, 1,000+ PRs total

Widsley Inc.:
- High-difficulty project including communication control
- QA follows detailed test steps and logs bugs with repro conditions to
  Red Codens
- Red Codens analyzes logs to identify the issue; Purple implements the
  fix
- PM aligns with the business side, drafts the PRD in Green Codens, and
  hands off to Purple
- Engineers step in via MCP when implementation gets stuck
- Result: 1 PM/engineer + 2 QA, 3 months, high-difficulty release shipped

### Media coverage

Codens press releases were republished via PR TIMES by 40+ Japanese
media outlets across two releases:
- Jul 15, 2026 release (22 pickups): Toyo Keizai Online, Asahi Shimbun
  Digital &[and], NewsPicks, and ASCII STARTUP, plus Jiji.com, NIKKEI
  COMPASS, Infoseek News, STRAIGHT PRESS, Otakuma Keizai Shimbun, and
  others
- Apr 30, 2026 release (25 pickups): overlapping set of outlets
Full press release listings: https://prtimes.jp/main/html/rd/p/000000008.000137480.html (Jul) and https://prtimes.jp/main/html/rd/p/000000007.000137480.html (Apr)

### Products (Codens Suite — 5 products + Auth)

- Green Codens — AI PRD creation & management. Converts free-form
  requests into structured PRDs, custom templates, review/approval flow,
  Notion/GitHub bidirectional sync. https://www.green.codens.ai
- Purple Codens — AI development orchestration. Detects Notion tasks,
  auto-launches an AI engineer, implements/tests/fixes autonomously, opens
  PRs. https://www.purple.codens.ai
- Orange Codens — AI code review & security audit. Auto-reviews GitHub
  PRs, detects bugs/vulnerabilities/design risk, SCA/SBOM dependency
  audit, hands findings to Purple/Red for automated fix PRs.
  https://www.orange.codens.ai
- Blue Codens — AI QA automation. Generates test cases from spec/code,
  generates Playwright E2E tests from natural language, analyzes test
  failures into bug reports, connects to Red Codens for the fix loop.
  https://www.blue.codens.ai
- Red Codens — AI automatic code repair. Ingests errors via webhook in
  real time, analyzes the whole codebase with Claude, retries up to 3x on
  test failure, opens a GitHub PR + Slack notification. 3-minute fix
  turnaround, 92% success rate. https://www.red.codens.ai
- Auth Codens — unified OAuth2/OIDC authentication for the whole family.
  https://auth.codens.ai

### Integrations

GitHub, Slack, Notion, Jira, Sentry

### Enterprise onboarding

Tier 4 — hands-on onboarding plan, from ¥1,000,000/month. Dedicated
onboarding team pairs with your team for a 3-6 month rollout: PRD
authoring, workflow design, continuous prompt/instruction tuning, internal
AI operating guidelines, weekly reviews.

### Common concerns, answered (safety FAQ)

Q: Will AI merge broken code directly to main?
A: Per-workflow you choose "wait for human merge" (default) or
"auto-merge". By default AI only opens a GitHub PR; main never changes
until you click merge. Even with auto-merge on low-risk/sandbox projects,
CI and the verify chain (implement -> test -> fix, 3-retry loop) always
run, and a PR that fails tests is never merged.

Q: Could AI run away and burn through credits?
A: Every task's credit consumption is logged in the Billing Control Plane
and viewable in real time on the dashboard. The fix_verify loop has
per-model retry caps by default (e.g. 3 for Claude-family, 6 for
Qwen-family), so infinite loops can't happen structurally. Monthly caps
per project are also configurable.

Q: Will I lose track of what AI did behind the scenes?
A: Purple Codens' workflow execution log records each step (analyze,
implement, verify, retry, escalation, merge decision) in chronological
order, including which model ran what and where it changed course. Every
PR's diff is visible and commentable on GitHub as usual.

Q: Could AI-generated code violate business rules?
A: Put CLAUDE.md / .claude/rules/ in the repo and Codens always includes
them in its prompts. For anything machine-checkable (naming conventions,
import order, banned APIs), a separate deterministic-enforcement layer
blocks AI output before merge, independent of the AI's own judgment.

Q: Will credentials or production data be handed directly to AI?
A: API keys and DB connection info live in AWS Secrets Manager; each task
accesses them only via a task-scoped IAM Role. AI never receives plaintext
credentials. Destructive production DB operations (DROP/TRUNCATE/mass
DELETE) are guarded at the workflow level and require explicit
authorization.

Q: If AI breaks something, can we roll back?
A: Every change is a GitHub PR/commit, so a normal `git revert` restores
the previous state. During the 14-day free trial you can use a sandbox
repo instead of production, and a Customer Success Manager helps design a
limited initial scope for the first project.

================================================================================
## PRICING (https://www.codens.ai/pricing.html)
================================================================================

1 credit = ¥0.01. Input token = 0.0225 credits, output token = 0.1125
credits (rate card effective July 2026), with a per-model multiplier
(default in-house Qwen 0.1x, Composer 0.5x, Claude Sonnet 5 / Haiku 4.5
1.0x, Claude Sonnet 4.6 3.0x, Claude Opus 4.7/4.8 5.0x, Fugu 5.0x, Claude
Fable 5 10.0x). Consumption varies by task complexity and model choice;
as one real example, Orange's PR review costs a median of ~79 credits
(~¥0.8, p90 ~260 credits/~¥2.6), based on our own production usage as of
July 2026.

- Free Trial — ¥0 / 14 days, no credit card. 100,000 credits.
- Hobby — ¥3,000/month. 300,000 credits/month. Red, Blue, Green Codens.
- Pro — ¥10,000/month. 1,000,000 credits/month. Red, Blue, Green, Orange
  Codens.
- Business — ¥50,000/month. 5,000,000 credits/month. Full suite, SLA
  guaranteed.
- Enterprise — custom contract. Unlimited credits, on-premises option.
- Onboarding Tier 4 (hands-on) — from ¥1,000,000/month, 3-month minimum
  contract, separate from the SaaS plan pricing above.

Annual billing: 20% discount vs. monthly (e.g. Business ¥50,000/mo ->
¥480,000/yr, i.e. ¥40,000/mo equivalent). Invoice billing available from
Pro tier up. Credits are shared across all products within a plan. Top-up
packs: ¥1,000 for 100,000 / ¥5,000 for 525,000 (+5%) / ¥10,000 for
1,100,000 (+10%) / ¥50,000 for 5,750,000 (+15%).

================================================================================
## COMPARISON (https://www.codens.ai/comparison.html)
================================================================================

Selection criteria: the four compared tools each represent the most
commonly compared tool in one of the four phases Codens unifies into a
single suite — "AI development agent" (Devin), "code completion" (GitHub
Copilot), "AI-native IDE" (Cursor), "AI PR review" (CodeRabbit). Pricing
and features reflect publicly available information as of July 2026,
researched by Codens; always verify against each vendor's official site.

### vs Devin (Cognition AI)

Devin's strengths: broad task range as an autonomous agent; strong track
record and name recognition in English-speaking markets.
Devin's weaknesses: no production monitoring or webhook-driven auto-fix;
high price ($500/user/month); no Japanese language or market support.
Verdict: Devin acts on instructions; Codens' Red Codens triggers
automatically from production monitoring and runs an unattended detect ->
fix PR -> test loop, and Codens covers requirements/review/QA/analytics
as one suite, with native Japanese support and JPY billing.

### vs GitHub Copilot

Copilot's strengths: best-in-class in-editor completion; deep
GitHub/Microsoft ecosystem integration; massive adoption.
Copilot's weaknesses: stops at completion — no auto-fix or test
generation; PR review is reviewer-requested with weak automation.
Verdict: different roles, not competitive; Copilot speeds up writing
code, Codens automates what happens to code after it's written (fixing,
reviewing, testing). Often used together.

### vs Cursor

Cursor's strengths: strong IDE experience and context understanding;
BugBot auto-merges ~35% of its findings (publicly reported), a genuinely
advanced approach.
Cursor's weaknesses: requires the Cursor IDE (hard to reach GitHub-only
teams); no CI/CD, production-monitoring, or test-automation integration.
Verdict: Codens works with any IDE via GitHub and automates the whole
team's workflow, not just the editor.

### vs CodeRabbit

CodeRabbit's strengths: deployed on 2M+ repositories, dominant share of
AI PR review; 40+ linters keep findings low-noise (low false-positive
rate); IDE/CLI/PR three-surface coverage; free for public repos.
CodeRabbit's weaknesses: stops at commenting — developers fix issues
manually; no way to reconcile code against a PRD; English only, no
Japanese comments, USD billing.
Verdict: Codens' Orange Codens review connects directly to Purple
Codens' automatic fix-PR generation in one product, with Japanese
comments and JPY billing.

### Comparison table summary

| Dimension                         | Codens Suite | Devin  | Copilot | Cursor | CodeRabbit |
|------------------------------------|--------------|--------|---------|--------|------------|
| Core role                          | whole lifecycle | autonomous agent | code completion | AI-native IDE | AI PR review |
| Production error auto-detect/fix   | yes (Sentry-integrated) | partial (instruction-based) | no | no | no |
| PR review + auto fix-PR            | yes | partial | partial (reviewer-requested) | partial (BugBot) | partial (comments only) |
| In-editor code completion          | partial | partial | yes (best-in-class) | yes | no |
| Automatic E2E test generation      | yes (Playwright) | partial | no | no | no |
| PRD/requirements alignment         | yes (Green Codens) | no | no | no | no |
| Native Japanese / JPY billing      | yes | no | partial | partial | no |
| Price (approx., monthly)           | from ¥3,000 | $500/user | from $19/user | from $40/user | $24/dev (Pro) |

================================================================================
## FAQ (site-wide, matches visible FAQPage JSON-LD)
================================================================================

Q: What is Codens?
A: Codens is an AI development suite from Japan that covers the entire
software development lifecycle. Five integrated products — PRD generation
(Green), development orchestration (Purple), code review & security audit
(Orange), test automation (Blue), and automatic error repair (Red) — let
teams build and operate software products without a large team of
specialist engineers, cutting lead time by up to 80% and bug recurrence
by up to 90%.

Q: Which products are included?
A: Green (PRD generation), Purple (development orchestration), Orange
(code review & security audit), Blue (test automation), and Red
(automatic error repair), plus the unified Auth Codens identity provider.
Each product works standalone or together as a suite.

Q: How do AI agents use Codens?
A: MCP-compatible agents such as Claude Code, Cursor, and Cline call 30
tools through a single MCP server (codens-mcp). Agents authenticate with
the Device Code Flow (RFC 8628) and run implementation, testing, repair,
and PRD generation without any UI navigation.

Q: How does pricing work?
A: Usage-based credit pricing (1 credit = JPY 0.01). Billed by task
complexity and token usage; credits are shared across all products within
a plan. New organizations receive a credit grant on signup.

Q: Is my data secure?
A: Codens uses tenant isolation, least-privilege GitHub App scoping, the
Device Code Flow, and secure secret management. See the Security FAQ
(https://www.codens.ai/security-faq.html) for details.

Q: Can I use Codens without engineers?
A: Yes. In the Portament case study, a team of 3 business members plus 1
engineer ran 5 projects in parallel and shipped 1,000+ PRs in 1.5 months.
Green Codens structures requests into a PRD, Purple Codens implements
them, and engineers are looped in only when needed. Teams can build and
operate software products without a large team of specialist engineers.

Q: How is Codens different from Devin?
A: See Comparison section above.

Q: How is Codens different from GitHub Copilot?
A: See Comparison section above.

Q: Can AI break production?
A: By default, AI only opens a PR — main never changes until a human
clicks merge. Even with auto-merge enabled, CI and a verify chain
(implement -> test -> fix, up to 3 retries) always run, so a PR that
fails tests is never merged. Credit consumption has retry caps to prevent
runaway loops, and every change lives as a GitHub PR/commit, so you can
always git revert.

Q: What do I need to get started?
A: Just an existing GitHub repository — setup takes as little as 30
minutes. Start a 14-day free trial with no credit card required, and use
Codens through MCP-compatible agents like Claude Code, Cursor, or Cline,
or directly from the browser. For larger rollouts, a dedicated onboarding
plan is also available.

================================================================================
## LINKS
================================================================================

- Home (JA): https://www.codens.ai/
- Home (EN): https://www.codens.ai/en/
- Pricing: https://www.codens.ai/pricing.html
- Comparison: https://www.codens.ai/comparison.html
- Stats dashboard: https://www.codens.ai/stats/
- Security FAQ: https://www.codens.ai/security-faq.html
- Contact: https://www.codens.ai/contact.html
- Help / Agent Reference: https://help.codens.ai/
- Help full text: https://help.codens.ai/llms-full.txt
- Short overview: https://www.codens.ai/llms.txt