# T25 Factory

> T25 Factory is a harness that runs locally for the coding agents a team already pays for, not another agent. It takes a task written in plain language and drives it through spec, plan, a human approval gate, implementation, QA, review and security checks until it opens a pull request. The code is written by those CLIs (Claude Code, Codex, Cursor, Gemini, Kimi, OpenCode, Grok, Pi, Vibe); T25 owns the policy: isolated git worktree per task, deterministic review evaluation, diff limits, and a merge that stays human.

The landing page is published at https://t25.io/ (Brazilian Portuguese) and https://t25.io/en/ (English). The blog is published in English and Brazilian Portuguese; every post has a clean Markdown copy at `<post-url>/index.md`.

## Product

- [Landing page (pt-BR)](https://t25.io/): what T25 does, the task path, FAQ and install commands
- [Landing page (English)](https://t25.io/en/): product path, FAQ and install commands
- [Documentation](https://t25.io/docs/): mental model, guides, architecture, configuration, CLI and API reference
- [Documentation in English](https://t25.io/en/docs/): static English documentation pages
- [License](https://t25.io/en/license/): full text of the evaluation license

## Blog (English)

- [What is a software factory for AI coding agents (and what it isn't)](https://t25.io/en/blog/what-is-a-software-factory-for-ai-coding-agents/index.md): A software factory for AI coding agents takes a task from spec to pull request through fixed states, approval gates, isolated git worktrees and QA. How it works, and how it differs from running one agent on its own.
- [The model's APPROVE is not the merge](https://t25.io/en/blog/the-models-approve-is-not-the-merge/index.md): An AI code review agent writes APPROVE, but who decides the merge? How T25 recomputes the verdict from findings, the bug that taught us to, and how to build a human gate that doesn't depend on the model.
- [One task, one worktree: isolating AI coding agents with git worktree](https://t25.io/en/blog/one-task-one-worktree-isolating-ai-coding-agents/index.md): How to use git worktree to isolate AI coding agents like Claude Code and Codex, one task per worktree. The path, ownership, lock and cleanup guards T25 uses, and the mistakes they prevent.

## Documentation (English)

- [index](https://t25.io/en/docs/)
- [architecture](https://t25.io/en/docs/architecture.html)
- [configuration](https://t25.io/en/docs/configuration.html)
- [api-reference](https://t25.io/en/docs/api-reference.html)
- [specifications](https://t25.io/en/docs/specifications.html)
- [guides/index](https://t25.io/en/docs/guides/)
- [guides/quickstart](https://t25.io/en/docs/guides/quickstart.html)
- [cli](https://t25.io/en/docs/cli.html)

## Blog (Português)

- [O que é uma software factory para agentes de código (e o que ela não é)](https://t25.io/blog/o-que-e-uma-software-factory-para-agentes-de-codigo/index.md): Software factory para agentes de código é a camada que leva uma tarefa da spec ao pull request com estados, gates, worktree isolado e QA. Veja como funciona e onde ela difere de um agente sozinho.
- [O APPROVE do modelo não é o merge](https://t25.io/blog/o-approve-do-modelo-nao-e-o-merge/index.md): Um agente de code review escreve APPROVE, mas quem decide o merge? Veja como o T25 recalcula o veredito a partir dos findings, o bug que nos ensinou isso e como montar um gate humano que não depende do modelo.
- [Uma ordem, um worktree: isolando agentes de código com git worktree](https://t25.io/blog/uma-ordem-um-worktree-isolando-agentes-de-codigo/index.md): Como usar git worktree para isolar agentes de código como Claude Code e Codex, uma tarefa por worktree. Os guards de caminho, dono, lock e limpeza que o T25 usa, e os erros que eles evitam.

## Optional

- [Full text of every blog post](https://t25.io/llms-full.txt)
- [RSS (English)](https://t25.io/en/blog/rss.xml)
- [RSS (Português)](https://t25.io/blog/rss.xml)