Forgenta Documentation
Forgenta is an AI-powered software studio with 30 specialized agents that design, develop, test, and deploy your software product autonomously. This documentation covers everything you need to use Forgenta effectively.
What Forgenta delivers
- Complete source code committed to your GitHub repository
- Infrastructure configured (VPS, domain, SSL, database)
- All services running in production
- SEO optimized (meta tags, sitemap, structured data)
- Project documentation and deployment report
Quickstart
1. Create your account
You need an invitation code to register. Once you have one, go to /auth and register with your email, full name, and a secure password.
2. Start a new project
From the dashboard, click New Project and describe your idea. Be as specific as possible:
I want to build a SaaS app for freelance designers. Users can create portfolios, receive client briefs, submit proposals, and get paid via Stripe. Language: English. Target: web + mobile (React Native).
3. Approve the strategy
The AI CEO will analyze your request and return a structured strategy document within minutes. Review and approve (or request modifications) via the dashboard or Telegram.
4. Track progress
Once approved, the agents activate. Monitor live progress in the dashboard — agent activity, completion percentage, and milestone updates appear in real time.
5. Receive your product
When all agents complete their work, you receive a final report with your GitHub repository link, live deployment URL, and all credentials.
Telegram Bot
Forgenta can be fully operated via Telegram — including project creation, approvals, and progress monitoring.
Getting started with Telegram
- Search for
@forgenta_boton Telegram - Send
/startand follow the authentication flow - Send your project idea as a text or voice message
- Approve milestones when the bot prompts you
Voice messages
Forgenta uses OpenAI Whisper to transcribe your voice messages. You can describe your project idea by speaking — in any supported language. The transcription is processed and routed to the appropriate agents automatically.
Supported Languages
All 30 agents strictly respond in the language you select. Every generated document, code comment, variable name description, and communication will be in your chosen language.
| Language | Code | Notes |
|---|---|---|
| English | en | Default |
| French | fr | Full support |
| Arabic (Standard) | ar | Modern Standard Arabic |
| Moroccan Darija | ar-MA | Colloquial Moroccan Arabic |
| Egyptian Arabic | ar-EG | Colloquial Egyptian Arabic |
| Spanish | es | Full support |
| Portuguese | pt | Full support |
Language is detected automatically from your input and can also be set explicitly in the New Project form under Project Language.
How It Works
Forgenta uses a two-phase LangGraph orchestration pipeline:
Phase 1: Briefing Graph
The CEO Agent receives your raw input and activates the Board (CTO, CPO, CFO, CMO, CSO). Each executive agent contributes their domain perspective to produce a comprehensive project brief — including tech stack recommendations, scope definition, risk analysis, marketing strategy, and financial model.
Phase 2: Execution Graph
After your approval, the Orchestrator Agent activates the 23 remaining agents in the correct order. Each agent reads the project state, performs its specialized task, and writes results back to the shared project state. Agents communicate asynchronously through a PostgreSQL + Redis memory layer.
Evolution Layer
Separately from active projects, three support agents run on a nightly schedule: Research Agent (discovers new technologies and best practices), Self-Improvement Agent (analyzes and refines all agent code), and Memory Agent (maintains the knowledge base).
Project Lifecycle
| Status | Description |
|---|---|
pending | Project received, Board analysis in progress |
awaiting_approval | Strategy ready — waiting for your approval |
in_progress | Agents actively building your project |
completed | All agents finished, deliverables ready |
failed | An error occurred — contact support |
Approval Flow
Before executing any major phase, the AI CEO presents a strategy for your review. You have three options:
- Approve — Agents proceed immediately with the current strategy
- Modify — You provide notes; the Board revises the strategy
- Reject — Project is cancelled and archived
Approvals can be done via the dashboard or by replying to the Telegram bot prompt.
File Uploads
You can attach files to any project for additional context:
- Design files (PNG, JPG) — UI/UX agents use these as visual references
- Documents (PDF, DOCX) — Analyzed for requirements and specifications
- Existing code (ZIP, individual files) — Used for the Improve Project feature
- Audio (WebM, MP3, WAV, M4A) — Transcribed via Whisper and processed as text
Maximum file sizes: 25MB for audio, 250MB for all other files.
Direction Agents (7)
The executive layer coordinates the entire project and makes strategic decisions.
| Agent | Model | Responsibility |
|---|---|---|
| CEO | Claude Opus 4.6 | Master coordinator. Receives input, delegates to Board, validates deliverables, maintains project vision throughout all phases. |
| CTO | Claude Sonnet 4.6 | Technology stack selection, architecture review, technical risk assessment, infrastructure strategy. |
| CPO | Claude Sonnet 4.6 | Product definition, user story creation, feature prioritization, roadmap planning. |
| CFO | Claude Sonnet 4.6 | Cost estimation, monetization strategy, financial modeling, pricing recommendations. |
| CMO | Claude Sonnet 4.6 | Go-to-market strategy, positioning, messaging, marketing plan, launch checklist. |
| CSO | Claude Sonnet 4.6 | Security policy, compliance requirements, data protection strategy, threat modeling. |
| Orchestrator | Claude Opus 4.6 | Activates and sequences all execution-phase agents, monitors progress, handles inter-agent dependencies. |
Product Agents (4)
| Agent | Model | Responsibility |
|---|---|---|
| Analyst | Claude Sonnet 4.6 | Market analysis, competitor research, requirements gathering, success metrics definition. |
| Product Manager | Claude Sonnet 4.6 | Detailed specifications, sprint planning, feature breakdown, acceptance criteria. |
| Architect | Claude Opus 4.6 | System design, database schema, API contracts, service boundaries, scalability planning. |
| UI/UX Designer | Claude Sonnet 4.6 | Interface design, component library, user flows, accessibility guidelines, design system tokens. |
Development Agents (5)
| Agent | Model | Responsibility |
|---|---|---|
| Frontend Developer | Claude Sonnet 4.6 | React/Next.js or Vue UI implementation, component development, responsive design, state management. |
| Backend Developer | Claude Opus 4.6 | API development (FastAPI/Django/Node.js), business logic, authentication, database integration. |
| Mobile Developer | Claude Sonnet 4.6 | React Native or Flutter app development, platform-specific features, App Store configuration. |
| Database Agent | Claude Sonnet 4.6 | Schema design, migrations, query optimization, indexing strategy, backup configuration. |
| API Integrator | Claude Sonnet 4.6 | Third-party API integrations (Stripe, SendGrid, Twilio, etc.), webhook setup, OAuth flows. |
Quality Agents (5)
| Agent | Model | Responsibility |
|---|---|---|
| Code Reviewer | Claude Opus 4.6 | Static analysis, code quality review, pattern enforcement, refactoring recommendations. |
| Security Agent | Claude Opus 4.6 | Vulnerability scanning (OWASP Top 10), dependency audit, auth hardening, data validation. |
| QA Tester | Claude Sonnet 4.6 | Test case generation, unit/integration/E2E test writing, coverage reporting. |
| Performance Agent | Claude Sonnet 4.6 | Load time optimization, database query tuning, caching strategy, bundle size analysis. |
| Bug Fixer | Claude Sonnet 4.6 | Error analysis, root cause identification, automated fix implementation, regression testing. |
DevOps Agents (6)
| Agent | Model | Responsibility |
|---|---|---|
| DevOps Agent | Claude Sonnet 4.6 | Server provisioning, Docker configuration, CI/CD pipeline setup (GitHub Actions), systemd services. |
| Domain Agent | Claude Sonnet 4.6 | DNS configuration, SSL certificate issuance (Let's Encrypt), subdomain management. |
| Cloudflare Agent | Claude Sonnet 4.6 | Cloudflare zone configuration, WAF rules, caching policies, DDoS protection settings. |
| SEO Agent | Claude Sonnet 4.6 | Meta tag generation, structured data (JSON-LD), sitemap creation, Core Web Vitals optimization. |
| Marketing Agent | Claude Sonnet 4.6 | Copy writing, email campaigns, social media content, launch materials. |
| Report Agent | Claude Sonnet 4.6 | Final project report, deployment summary, handover documentation, credentials package. |
Support Agents (3)
| Agent | Model | Responsibility |
|---|---|---|
| Memory Agent | Claude Haiku 4.5 | Knowledge base maintenance, cross-project learning, context retrieval, memory indexing. |
| Research Agent | Gemini Pro | Nightly web research on new frameworks, AI model updates, security advisories, and industry trends. |
| Self-Improvement | Claude Opus 4.6 | Nightly review and enhancement of all agent system prompts, code, and decision logic based on recent performance. |
API — Authentication
Base URL: https://api.forgenta.ai (or https://forgenta.ai/api via nginx proxy).
All protected endpoints require a Bearer token in the Authorization header:
Authorization: Bearer <your-jwt-token>
| Method | Endpoint | Description |
|---|---|---|
| POST | /auth/login |
Authenticate and receive a JWT token. Body: {"email": "...", "password": "..."} |
| POST | /auth/register |
Create account with invitation code. Body: {"full_name", "email", "password", "invitation_code"} |
| GET | /auth/me |
Return current user profile. Requires token. |
| PATCH | /auth/profile |
Update profile (full_name, email, password). Requires token. |
| GET | /auth/settings |
Get notification and preference settings. Requires token. |
| PATCH | /auth/settings |
Update settings. Body: {"notifications_email": true, "language": "fr", ...} |
API — Projects
| Method | Endpoint | Description |
|---|---|---|
| GET | /projects/v2 |
List your projects. Query params: page, limit, status |
| POST | /projects/v2 |
Create a new project. Body: {"raw_input": "...", "language": "fr", "project_type": "saas"} |
| GET | /projects/v2/{id} |
Get project details, status, agent logs, and live progress. |
| POST | /projects/v2/{id}/validate |
Approve/reject a strategy. Body: {"action": "approve|reject|modify", "notes": "..."} |
| DELETE | /projects/v2/{id} |
Hard-delete a project and all associated data. |
| GET | /projects/v2/{id}/live |
Real-time status: progress percentage, current agent, ETA, recent logs. |
API — Uploads
| Method | Endpoint | Description |
|---|---|---|
| POST | /upload/transcribe |
Transcribe an audio file via Whisper. Multipart form field: file. Returns {"text": "..."}. Max 25MB. |
| POST | /upload/file |
Generic file upload. Multipart form field: file. Max 250MB. Returns file metadata and path. |
API — Scheduler
| Method | Endpoint | Description |
|---|---|---|
| POST | /scheduler/{job_name}/run |
Manually trigger a scheduled job. Job names: research, self_improvement, memory_cleanup |
API — Admin (Superadmin only)
| Method | Endpoint | Description |
|---|---|---|
| GET | /admin/users |
List all users with pagination. |
| POST | /admin/users |
Create a new user manually (no invitation required). |
| POST | /admin/invitations |
Generate invitation codes. Body: {"count": 5, "expires_days": 30} |
| POST | /admin/research/run |
Trigger the Research Agent immediately (non-blocking). |
| POST | /admin/self-improvement/run |
Trigger the Self-Improvement Agent immediately (non-blocking). |
| GET | /admin/health |
System health: PostgreSQL, Redis, agent status, disk, memory, CPU. |
FAQ
Can I use Forgenta to improve existing code?
Yes. Use the Improve Project feature in the dashboard. Upload your project as a ZIP file or select an existing Forgenta project. Describe the improvements needed and the agents will analyze and enhance your codebase.
What happens if an agent fails mid-project?
The Orchestrator Agent monitors all agent executions. If an agent fails, it retries up to 3 times with exponential backoff. If it still fails, the Bug Fixer Agent is activated. You are notified via the dashboard and Telegram if manual intervention is needed.
Can I run Forgenta on my own infrastructure?
Self-hosted deployment is planned for a future release. Currently, Forgenta runs on managed infrastructure. The source code for the core platform is not yet open-source.
How do I connect my own server and domain?
During project creation, you can provide your server credentials (IP, SSH key) and domain name. The DevOps and Domain agents will configure everything automatically. Cloudflare integration is supported natively.
Is there an API rate limit?
Yes. Standard beta users: 10 concurrent agent requests, 50 project creations per month. Rate limits are subject to change during beta. Contact us if you need higher limits.
Changelog
v3.0 — March 2026
- Strict language enforcement across all 30 agents (no language switching)
- Hard delete for projects (immediate removal from DB and UI)
- Voice input via Whisper API (MediaRecorder → transcription)
- URL routing with History API in the dashboard (deep linking)
- Settings persistence via
/auth/settingswith JSONB storage - Session detection on landing page (auto-redirect for logged-in users)
- Project detail: Preview, Download ZIP, Mini Chat with file upload
- Improve Project redesign with source selector and credentials panel
- Avatar dropdown: View Profile, Edit Profile, Sign Out
- Research + Self-Improvement "Run Now" buttons in admin panel
v2.0 — February 2026
- Full violet dark design system applied across all pages
- JWT authentication system with invitation codes
- PostgreSQL integration with SQLAlchemy ORM
- File upload endpoint (250MB) with automatic cleanup
- HTTPS/SSL via Certbot (Let's Encrypt)
- Live project tracking with real-time agent logs
- Admin panel: users, invitations, system health
v1.0 — January 2026
- Initial 27-agent system (CEO, Board, 5 dev teams)
- Telegram bot integration
- LangGraph orchestration (briefing + execution graphs)
- Multi-language support (EN, FR, AR)
- Evolution Scheduler (research + self-improvement)