Corevice Launches "Codens," a 5-Agent Development Suite for Fully Automated AI Development and Operations & Maintenance; Users Can Develop Full-Fledged Systems with Only Ideas and Judgments
Corevice launched "Codens," a 5-agent development suite that fully automates AI development and operations & maintenance, on April 29, 2026. It features the industry's first fully transparent per-token pricing model, allowing users to calculate costs in advance.
📋 Article Processing Timeline
- 📰 Published: April 30, 2026 at 19:21
- 🔍 Collected: April 30, 2026 at 11:01
- 🤖 AI Analyzed: April 30, 2026 at 11:17 (15 min after Collected)
"From AI that 'writes' to AI that 'creates'."
Corevice Inc. (Headquarters: Osaka, Representative: Takayuki Kawazoe, hereinafter "Corevice") officially launched "Codens," an AI development automation suite, on April 29, 2026. Codens is a subscription-based SaaS composed of five specialized AI agents covering five areas of the software development lifecycle (PRD creation, workflow instruction, automatic error correction, test automation, and engineering activity visualization), sharing a common credit pool within an organization.
The biggest differentiator is the full disclosure of its per-token rate card. While many AI development tools present prices in units like "N cases per month" or "credit system," Codens discloses the yen conversion per credit, token conversion formula, and model-specific coefficients, allowing users to calculate the cost of any operation before signing a contract.
LP: https://www.codens.ai/
■ Summary
SaaS "Codens," composed of five AI agents sharing a common credit pool within an organization, launched on April 29, 2026.
Industry-first, fully transparent per-token rate card (1 credit = ¥0.01, model coefficients: Haiku 0.2x / Sonnet 1.0x / Opus 5.0x).
Automates developers' manual tasks in five areas, including automatic correction of production errors with Sentry, automatic generation of E2E tests, automatic structuring of PRDs, and automatic visualization of engineering activities.
Hobby ¥3,000/month to Business ¥50,000/month, 14-day free trial (30,000 credits granted, no credit card required).
■ Background: The "Price Black Box" Problem of AI Development Tools
In recent years, AI-powered development support tools have rapidly become widespread. However, the challenge Corevice's representative, Kawazoe, felt on the ground was that "users cannot see what and how much is being consumed."
Most major competing products charge a fixed monthly fee based on abstract units like "N requests" or "N tasks." How many tokens one request consumes behind the scenes, which model is called, and how much markup is applied to the actual payment to the API provider (Anthropic / OpenAI, etc.) – these are not disclosed.
As a result, users are in a situation where:
ROI cannot be estimated (actual cost per task is unknown).
Products cannot be compared (units vary, making quantitative comparison impossible).
Monthly budget cannot be predicted (relationship between usage and cost is unclear).
Corevice believes that "the ability for users to calculate fees is a basic requirement for AI development tools" and has designed Codens with rate card disclosure in mind from the outset.
■ Codens' Value Proposition: Full Disclosure of Per-Token Rate Card
Codens publicly discloses the following pricing formula as its pricing rationale:
1 credit = ¥0.01 (≒ $0.000067)
1 input token = 1.5 credits × model coefficient
1 output token = 7.5 credits × model coefficient
Model coefficients = Haiku 0.2x / Sonnet 1.0x / Opus 5.0x
For example, a typical task processing 15,000 input tokens and 8,000 output tokens with Sonnet can be calculated by the user before the contract as "15,000 × 1.5 + 8,000 × 7.5 = 82,500 credits = ¥825." By multiplying by the number of retries and integration overhead, the monthly budget can also be designed bottom-up.
Retail prices are designed top-down with a target gross profit margin of approximately 80%, which is standard in the SaaS industry, and the above per-token unit price is derived as a result.
■ Five Specialized AI Agents Comprising Codens
Agent Name | Area | Main Functions
---|---|---
Green Codens | PRD Creation/Management | AI structuring of business requirements → Notion sync, review/approval flow, GitHub integration
Purple Codens | Workflow Instruction | Automatic decomposition of PRD into implementation tasks, Notion ticket-driven, agent call control
Corevice Inc. (Headquarters: Osaka, Representative: Takayuki Kawazoe, hereinafter "Corevice") officially launched "Codens," an AI development automation suite, on April 29, 2026. Codens is a subscription-based SaaS composed of five specialized AI agents covering five areas of the software development lifecycle (PRD creation, workflow instruction, automatic error correction, test automation, and engineering activity visualization), sharing a common credit pool within an organization.
The biggest differentiator is the full disclosure of its per-token rate card. While many AI development tools present prices in units like "N cases per month" or "credit system," Codens discloses the yen conversion per credit, token conversion formula, and model-specific coefficients, allowing users to calculate the cost of any operation before signing a contract.
LP: https://www.codens.ai/
■ Summary
SaaS "Codens," composed of five AI agents sharing a common credit pool within an organization, launched on April 29, 2026.
Industry-first, fully transparent per-token rate card (1 credit = ¥0.01, model coefficients: Haiku 0.2x / Sonnet 1.0x / Opus 5.0x).
Automates developers' manual tasks in five areas, including automatic correction of production errors with Sentry, automatic generation of E2E tests, automatic structuring of PRDs, and automatic visualization of engineering activities.
Hobby ¥3,000/month to Business ¥50,000/month, 14-day free trial (30,000 credits granted, no credit card required).
■ Background: The "Price Black Box" Problem of AI Development Tools
In recent years, AI-powered development support tools have rapidly become widespread. However, the challenge Corevice's representative, Kawazoe, felt on the ground was that "users cannot see what and how much is being consumed."
Most major competing products charge a fixed monthly fee based on abstract units like "N requests" or "N tasks." How many tokens one request consumes behind the scenes, which model is called, and how much markup is applied to the actual payment to the API provider (Anthropic / OpenAI, etc.) – these are not disclosed.
As a result, users are in a situation where:
ROI cannot be estimated (actual cost per task is unknown).
Products cannot be compared (units vary, making quantitative comparison impossible).
Monthly budget cannot be predicted (relationship between usage and cost is unclear).
Corevice believes that "the ability for users to calculate fees is a basic requirement for AI development tools" and has designed Codens with rate card disclosure in mind from the outset.
■ Codens' Value Proposition: Full Disclosure of Per-Token Rate Card
Codens publicly discloses the following pricing formula as its pricing rationale:
1 credit = ¥0.01 (≒ $0.000067)
1 input token = 1.5 credits × model coefficient
1 output token = 7.5 credits × model coefficient
Model coefficients = Haiku 0.2x / Sonnet 1.0x / Opus 5.0x
For example, a typical task processing 15,000 input tokens and 8,000 output tokens with Sonnet can be calculated by the user before the contract as "15,000 × 1.5 + 8,000 × 7.5 = 82,500 credits = ¥825." By multiplying by the number of retries and integration overhead, the monthly budget can also be designed bottom-up.
Retail prices are designed top-down with a target gross profit margin of approximately 80%, which is standard in the SaaS industry, and the above per-token unit price is derived as a result.
■ Five Specialized AI Agents Comprising Codens
Agent Name | Area | Main Functions
---|---|---
Green Codens | PRD Creation/Management | AI structuring of business requirements → Notion sync, review/approval flow, GitHub integration
Purple Codens | Workflow Instruction | Automatic decomposition of PRD into implementation tasks, Notion ticket-driven, agent call control