The technology behind Forge and Forenta Enterprise.

1. 01

   Idea

   Question

   What do we want to build, and why?

   Input

   Idea, note or problem statement

   Output

   Project brief

   Gate

   Goal, scope and assumptions captured

2. 02

   Research

   Question

   What do we already know, and what is missing?

   Input

   Project brief, sources and a repository (read-only)

   Output

   Research package

   Gate

   Evidence gathered and weighed

3. 03

   Analysis

   Question

   Does the rationale hold, and where is the risk?

   Input

   Research package and project context

   Output

   Advisory report

   Gate

   Findings assessed and an explicit verdict

4. 04

   Architecture

   Question

   How do we build this, and with what data model?

   Input

   Advisory report and requirements

   Output

   Blueprint

   Gate

   Design and execution plan complete

5. 05

   Build

   Question

   Does what we deliver actually work?

   Input

   Blueprint and build prompts

   Output

   Build report

   Gate

   Parts work and are traceable

6. 06

   Validation & security

   Question

   Is it correct, secure and ready?

   Input

   Build report and test criteria

   Output

   Test report

   Gate

   Tests, risks and a go/no-go decision

7. 07

   Deployment

   Question

   How do we ship this under control?

   Input

   Test report and release plan

   Output

   Release runbook

   Gate

   Release, rollback and checklist complete

8. 08

   Monitoring

   Question

   Does it do in production what it should?

   Input

   Release runbook and signals

   Output

   Monitoring report

   Gate

   Metrics and follow-up in place

1. 01Context
2. 02Analysis
3. 03Findings
4. 04Tasks
5. 05Checklist
6. 06Stage artifact
7. 07Build prompt
8. 08Gate

• Stage artifact

  One recorded document per stage: project brief, research package, advisory report, blueprint, build report, test report, release runbook or monitoring report.

  Advisory report · Analysis
  Verdict: conditional go
  Main risk: demand for the product untested

• Tasks and recommendations

  Concrete, traceable next steps that follow from the findings. Each task belongs to the stage and the artifact it came from.

  1. Test demand in 5 user interviews
  2. Define the data model for the core entity
  3. Cut scope to a single flow

• Checklist and gate

  A required checklist blocks the stage transition until every item is done. The server and the database enforce this. Overruling a gate is always explicit and recorded in the audit log.

  [x] Sources weighed
  [ ] Assumptions tested
  Gate blocked: 1 item open

• Build prompt

  An executable instruction with context and acceptance criteria, ready to paste into your own development environment. You get the task and its criteria, not internal system prompts.

  Build the input validation for the sign-up form.
  Acceptance: empty fields rejected, email checked, errors shown.

• Recorded decisions

  Decisions and stage transitions are recorded, so you can check what was decided and when.

• Gates that block

  Required checks stop an incomplete stage transition. The server and the database enforce this.

• Traceable findings

  Every finding can be traced to the tasks and artifacts it produces, across the stages.

• A person decides

  Important decisions are reviewed by a person. Overruling needs a reason and leaves the AI verdict as it was.

• GitHub

  Repository context, read-only: README, languages, branches, recent commits and open issues.

  Enterprise

• Jira and Confluence

  Recently updated issues and pages as project context, read-only.

  Enterprise

• Note or document

  Pasted text, such as architecture notes or decisions, as project context.

  Enterprise

• Slack

  Posts a project update to a channel when you ask. Reads nothing.

  Enterprise

• MCP sources

  Atlassian, Notion, Linear, Asana, GitHub, Sentry, Microsoft Learn or your own MCP server. A new source is read-only, masks email addresses, phone numbers, IBAN, BSN and card numbers, and has every tool off until an admin turns it on.

  Enterprise

• Claude Code and Codex

  With a personal organisation token, both tools read workspaces, projects, tasks and review findings. With write access they report work and evidence, handle findings, tick checks and request a review. Every call is recorded in the audit log.

  Enterprise

• Other AI tools

  Build prompts, checklists and artifacts as Markdown, to paste into your own tool.

  Via handoff

• Supabase and Vercel

  Schema, deployments and environment data as project context, read-only.

  In preparation

Where AI is used

Forge and Flux use Anthropic's commercial API. Optional avatars are generated through fal.ai. Reviews in Forenta Enterprise run on the organization's own model key, from Anthropic or OpenAI. That key is stored in Supabase Vault.

Which input is processed

Forge processes only the text you paste. In an Enterprise project, connected sources such as GitHub or Jira can serve as additional evidence.

Voice input

The microphone button uses your browser's speech recognition (the Web Speech API). Your browser does the recognition, not Forenta. Chrome sends the audio to Google's servers to do it, so what happens to your voice differs per browser. Forenta receives no audio, only the text you choose to keep. That text reaches Anthropic only when you ask Flux to structure it. If your browser does not support the API, the button does not appear.

Where processing happens

Core data is intended for an EU region of Supabase and the app runs on Vercel. Anthropic and fal.ai may process input in the United States, so not all processing stays in the EU. If you use the voice button, your browser's speech service is added to that.

Human control

Important decisions are reviewed by a person before they are made. Overruling always needs a reason, is recorded in the audit log and leaves the AI verdict as it was.

Retention and training

Forenta does not train its own models on your content. Anthropic states that commercial API data is not used for training by default and is normally removed from its backend within 30 days, subject to safety, legal and contractual exceptions.

AI can be wrong

AI output can contain errors and unfounded assumptions. Read it critically and use the reviews and gates to check it.

More on data processing and safeguards: privacy policy and trust and data.