文章目录步骤1. 登录你的 Reddit 账号2. 打开[Reddit Apps 控制台](https://www.reddit.com/prefs/apps?utm_sourcechatgpt.com)3. 滚动到底部点击create app 或 create another app4. 填写应用信息5. 创建后你会看到→问Claude←Claude回答但是现在似乎无法获取了步骤你需要在 Reddit Apps 控制台 创建一个应用才能获得 Reddit API credentials。步骤如下1. 登录你的 Reddit 账号2. 打开Reddit Apps 控制台3. 滚动到底部点击create app或create another app4. 填写应用信息name你的应用名称app typescript推荐适合个人脚本 / crawler / MVPweb appOAuth Webinstalled app桌面/移动端description可选about url可选redirect uriscript 类型可填http://localhost:8080web app 必须是真实 OAuth callback5. 创建后你会看到client id在应用名称下方那串短字符串client secretsecret那一行Python 示例适合你现在做 MVPimportpraw redditpraw.Reddit(client_idYOUR_CLIENT_ID,client_secretYOUR_CLIENT_SECRET,user_agentmy-mvp-app/0.1 by u/YOUR_USERNAME,)forpostinreddit.subreddit(Python).hot(limit5):print(post.title)需要安装pipinstallpraw其中client_idclient_secretuser_agent就是 Reddit API credentials 的核心。如果你正在做你那个 multi-agent 信息聚合 MVP我建议app type 直接选script使用PRAW简单或asyncpraw异步版本更适合 agent pipeline相关官方文档Reddit API DocsPRAW DocumentationAsync PRAW Documentation→问Claude我现在正在创建Reddit App但是它需要填写以下信息nameapplication typedescriptionabout urlredirect uri我该如何填写←Claude回答● 逐个填写 ┌──────────────────┬─────────────────────────────────────────────────────────────────────────┐ │ 字段 │ 填写内容 │ ├──────────────────┼─────────────────────────────────────────────────────────────────────────┤ │ name │ SignalForge │ ├──────────────────┼─────────────────────────────────────────────────────────────────────────┤ │ application type │ script用于个人脚本/bot不需要 OAuth 回调 │ ├──────────────────┼─────────────────────────────────────────────────────────────────────────┤ │ description │ 留空或写 Personal data collection script for SignalForge │ ├──────────────────┼─────────────────────────────────────────────────────────────────────────┤ │ about url │ 留空 │ ├──────────────────┼─────────────────────────────────────────────────────────────────────────┤ │ redirect uri │ http://localhost:8080 script 类型不会实际用到但是必填填这个就行 │ └──────────────────┴─────────────────────────────────────────────────────────────────────────┘ 关键是 application type 选 script这是 PRAW 个人脚本的标准类型不需要 web 回调流程直接用 client_id client_secret username password 认证。In order to create an application or use our API you can read our full policies here: https://support.reddithelp.com/hc/en-us/articles/42728983564564-Responsible-Builder-Policy但是现在似乎无法获取了✻ Churned for 6s How To Get (—Or Not—) Reddit API Credentials ● 这看起来像是一个页面标题或提示信息但内容没有完整发过来。你是想问什么还是想把某段说明贴给我看 ✻ Worked for 7s Current Reality Access Is Heavily Restricted: Reddit stopped self-service app creation and now requires manual approval, and many users report denials or silence after applying. Free access for personal/non-commercial use is still a thing though, with a 100 QPM rate limit per OAuth client — check their Responsible Builder Policy and apply at reddit.com/wiki/api Most Requests Are Rejected / No Reply: Multiple users say they applied and either got denied or never heard back. Yeah didnt got one either after waiting more than two weeks. They didnt even reply. This is ridiculous. Creation Is Effectively Blocked For New Apps: Community reports that creating new client IDs/apps is banned or broken; existing keys continue to work for now. Creating Client IDs has been banned by Reddit. ... Yes, its the creation of them that has been banned. Yours still functions for now. Process Is Flaky / Buggy: The dev app page can silently reload or fail; users suggest desktop/incognito, disabling extensions, sane redirect URIs. The dev app page is super flaky... use a desktop browser, disable extensions, try an incognito window, and switch networks... Fill every field, keep the redirect URI sane... and don’t mash create too fast. What To Try If You Need Credentials Follow Responsible Builder Policy Apply: Read and submit per the policy; reviewers expect detailed use-case information. In order to create an application or use our API you can read our full policies here: Responsible-Builder-Policy Be Ready To Justify Use Case: Projects that look like large-scale data extraction are likely to be rejected. Reddit isnt going to let you do this. They just wont approve your application. Escalate If Legitimate / Enterprise: Some approvals happen after escalation or via a Reddit sales contact for larger organizations. I did (for moderation tools), though I was originally denied too and only got my approval after escalating it. Workarounds For Now: Use public JSON/RSS endpoints or third‑party scraping/paid services if API access is unavailable. You can append .rss or .json to any subreddit or thread to get it in json and I made a free reddit scraper (no api key needed): ... Apify Consider Paid Pipes / Scrapers: Paid services (Apify, SerpApi, Scraping tools) and commercial social-listening platforms are commonly used. I’d lean on Pushshift replacements or paid pipes like SerpApi and Apify; for live monitoring ... people use ... Brandwatch and Sprinklr Practical Limits Tips Rate Limits Exist For Approved Clients: Approved non-commercial clients reportedly have a 100 QPM limit. Free access for personal/non-commercial use is still a thing though, with a 100 QPM rate limit per OAuth client Old Credentials Still Work — But May Be At Risk: Existing client IDs often continue functioning; some users saved old keys for apps. I found my old API key in emails for my deleted account... Using old key ... works for now. DMs / Conversations Are Not Publicly Exposed: Conversation/DM data is largely inaccessible via public JSON endpoints now. Reddit doesn’t expose [messages/conversations] in any public JSON endpoint anymore... a lot of the text ... is coming in via obfuscated GraphQL/batch endpoints Expect Slow Or No Replies: After submitting forms many report silence. usually you dont get an answer. Communities for Further Discussion ⎿ Interrupted · What should Claude do instead? reddit拷贝过来的现在是不可以了吗 ● 是的Reddit 现在基本封锁了新应用的注册。对 MVP 来说有两个务实的替代方案 方案 AReddit 公开 JSON 接口推荐 Reddit 每个页面加 .json 就能直接拿数据无需任何 API key https://www.reddit.com/r/python/top.json?limit100tweek - 无需认证无需 PRAW - 速率限制宽松每 2 秒一次即可 - 返回结构和 API 完全一样 - 需要把 Collector 里的 PRAW 换成 httpx已安装 方案 B合成 Mock 数据 写一个脚本生成仿真 Reddit 帖子直接注入 pipeline 跑端到端。 - 完全不依赖 Reddit - 适合重复测试但不是真实数据验证 --- 我的建议是方案 A改动只在 collector/reddit_client.py其他 agent 不动且验证效果更真实。 你倾向哪个方案
