# AIWriteX: Multi-Agent WeChat Article Automation with Trending Topic Detection

> AIWriteX is an Apache-licensed Python tool that uses a CrewAI multi-agent pipeline to automate the full WeChat public account workflow: it detects trending topics, generates and formats articles, applies anti-AI-detection processing, and publishes to WeChat and five other platforms.

**iniwap/AIWriteX** — AIWriteX - 微信公众号全自动AI工具：全网热搜舆情聚合+趋势分析+爆款选题+文章采集+一键生成排版发布 | AI自动配图 | 去AI味、过朱雀检测 | 支持小红书/百家号/头条等多平台 | 洗稿润色支持多账号 | 专家赛道 | 短视频文案 | 手机控制 | 小说连载 | 爆文10w+神器

- Repository: https://github.com/iniwap/AIWriteX
- Website: https://aiwritex.voidai.cc
- Stars: 2,042 · Forks: 409
- Language: Python
- License: Apache-2.0
- Published: 2026-09-10 · Updated: 2026-09-10 · Language: en
- Canonical page: https://hysenlabs.com/projects/iniwap-aiwritex

## What AIWriteX Is Built to Automate

AIWriteX targets WeChat public account operators who publish content regularly and want to reduce the manual work of finding topics, writing articles, sourcing images, formatting for WeChat's rich text editor, and hitting the publish button. The full pipeline covers each of those steps.

The tool supports six publishing platforms: WeChat public accounts (including scheduled and batch publishing, draft queue, and long-image mode), Xiaohongshu with image-text posts, Baijiahao, Toutiao, Weibo, and Fanqie Novel for serialised fiction. Each platform has a different content format, and the README describes per-platform support for drafts, auto-publishing, and long-image mode where the platform supports it.

Two modes are offered. Software mode provides a graphical interface intended for non-technical operators who want a visual workflow. Developer mode offers flexibility for those who want to modify the pipeline, connect additional models, or run the tool from the command line. Configuration is managed through two files: config.yaml for platform credentials, model settings, and publishing parameters, and aiforge.toml for the real-time search engine.

## The Multi-Agent Pipeline

AIWriteX uses the CrewAI framework to orchestrate a team of specialised agents on each article. The README lists four roles: a researcher who gathers real-time search results and reference articles, a writer who produces the draft, a reviewer who checks quality and applies corrections, and a designer who handles layout and formatting. The agents work in sequence, with each role's output passing to the next.

The AIForge Engine, listed as a required dependency in requirements.txt, provides the real-time search and article retrieval capability. Rather than generating articles from static training data, AIForge searches the web for current sources on the selected topic and feeds them into the writing pipeline. This is the mechanism behind the README's claim of high timeliness.

The trending topic module aggregates hot search data from Weibo, Douyin, Xiaohongshu, and other platforms. The README describes a prediction algorithm that claims to identify emerging topics two to six hours before they peak, based on momentum signals. The analysis feeds directly into the topic selection step of the writing workflow.

## Installing and Configuring AIWriteX

AIWriteX is a Python package requiring Python 3.10 to 3.12, as specified in pyproject.toml. The dependencies include CrewAI, aiforge-engine, pywebview for the GUI, FastAPI and uvicorn for the local web server, and Pillow for image processing. To install the dependencies:

```bash
pip install -r requirements.txt
```

The pyproject.toml defines an entry point that runs the application:

```bash
ai_write_x
```

Configuration requires editing config.yaml with API credentials. Key settings in the README include the platforms section for per-platform topic selection weights, the wechat section with appid, appsecret, and author for auto-publishing, and the api section with the AI model API key and api_type to select the model provider. The api_type can be set to OpenRouter or other compatible providers, and the model_index parameter selects the specific model within that provider. Image generation uses a separate img_api section supporting an Aliyun-backed model or a random Picsum image as a lower-cost alternative.

## Expert Tracks and the Copy Arsenal

Two features extend the core article generation to more specialised use cases. Expert tracks inject vertical domain expertise into the writing pipeline. Each track carries audience profile, content boundaries, structural rhythm, and quality standards that override the generic prompts. The README lists health, technology, news, entertainment, psychology, career development, culture, and history as built-in tracks. A track can have multiple template versions for different audience segments or platform styles, and the entire workflow, including prompts, default parameters, AI image generation, and anti-AI processing, is shaped by the active track.

The copy arsenal (文案武库) addresses short-form content such as video scripts, product promotion scripts, Xiaohongshu notes, and drama clips. The README lists 24 built-in content scenarios across 10 categories, each with a predefined structure, hook patterns, and tone guidelines. Three creation modes are available: original generation, imitation of an existing viral piece's structure, and conversion of existing content to a new format. Five adjustable parameters control hook strength, emotional intensity, pacing, closing style, and colloquial tone on a per-generation basis.

## The Anti-AI-Detection Feature

Chinese AI-detection tools such as Zhuque score articles for the probability that they were AI-generated. High AI scores can reduce the distribution of WeChat articles on the platform. AIWriteX includes processing it describes as a deep adversarial detection engine.

The README describes four techniques. Dynamic style mimicry extracts vocabulary patterns from a reference article and injects them into the generated text. Structure fragmentation breaks up AI-typical patterns by flattening lists, varying paragraph lengths, and removing connector words. Emotional enhancement injects subjective expressions and rhetorical questions to add apparent human bias. A scoring feedback loop lets operators rate each output to improve subsequent processing.

The README includes a candid assessment: the results are described as good but unstable, and the current implementation is called the strongest available but notes that Zhuque and similar detection tools continue to improve. The README explicitly recommends human review and editing rather than full AI generation, and calls the long-term effort against detection tools an ongoing process that requires sustained work.

## AIWriteX vs. WeWrite and Similar Tools

WeWrite is a tool for formatting Markdown content into WeChat's rich text editor style. The focus is on presentation: converting plain text with headings, code blocks, and images into a format that renders correctly in WeChat without losing formatting during copy-paste. AIWriteX does not compete on that specific formatter use case; it includes its own formatting as part of a larger automation pipeline.

The relevant difference is scope. WeWrite addresses a single step in the publishing workflow: formatting. AIWriteX attempts to automate the entire workflow from topic discovery through generation, formatting, anti-detection processing, and publishing. For operators who have a writing process and only need formatting help, WeWrite or Lyricat wechat-format is simpler. For operators who want to reduce the total time spent on the publishing cycle and are willing to review AI-generated drafts, AIWriteX covers ground that single-purpose formatters do not.

The multi-platform publishing support (Baijiahao, Toutiao, Weibo) also distinguishes AIWriteX from tools focused exclusively on WeChat. Operators running simultaneous accounts across several platforms are the primary audience for that feature.

## Conclusion

AIWriteX suits Chinese-language content operators who publish to WeChat, Xiaohongshu, Baijiahao, or Toutiao and want to reduce the manual work in each cycle. The anti-AI-detection feature is described as the strongest available but also as unstable, and the README explicitly recommends human review and editing over full AI generation. Developers who want to run the tool in software mode can use the GUI directly; those who want to extend or automate it further use developer mode with config.yaml and aiforge.toml. The last GitHub release was V2.5.1 on 2026-09-17 and the last push was on 2026-09-24.

## FAQ

### What AI models does AIWriteX support?

The config.yaml api section supports multiple model platforms via the api_type setting. The README specifically mentions OpenRouter and DeepSeek as configurable options. The model_index and api_key fields allow selection of a specific model and switching between multiple API keys to distribute load across free-tier quotas.

### Does AIWriteX require a WeChat official account to use?

WeChat auto-publishing requires an appid, appsecret, and author in the wechat section of config.yaml. The README notes that group-send functionality requires a verified (certified) account. Other platforms like Xiaohongshu and Baijiahao have their own credential requirements. Manual draft review is possible without auto-publish credentials.

### Can AIWriteX generate images automatically for articles?

Yes. The img_api section in config.yaml configures image generation. The ali option uses an Aliyun-backed image model and requires an api_key; the picsum option uses random stock images at no additional API cost. Version 2.5.1 added 16 built-in image tracks covering AI-generated and locally rendered images.

## Sources

- [iniwap/AIWriteX on GitHub](https://github.com/iniwap/AIWriteX)
- [License: Apache-2.0](https://github.com/iniwap/AIWriteX/blob/main/LICENSE)
- [Project website](https://aiwritex.voidai.cc)
- [README](https://github.com/iniwap/AIWriteX/blob/main/README.md)
- [Releases](https://github.com/iniwap/AIWriteX/releases)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/iniwap-aiwritex