agents-radar: A Daily AI Digest Pipeline That Publishes Bilingual Reports Automatically
Daily AI ecosystem digest from 10 sources (GitHub, ArXiv, HN, HuggingFace, Product Hunt, Dev.to, Lobste.rs). Bilingual ZH/EN reports via GitHub Actions.
At a glance
- What is it?
- agents-radar is a GitHub Actions workflow that aggregates AI ecosystem signals from 10 sources into bilingual daily digests, published as issues, Markdown files, RSS, and an MCP server. It is useful for developers who want a curated AI news feed without manual curation, but it has limitations in source coverage and customization.
- Who is it for?
- Adopt agents-radar if you want a zero-maintenance, automated daily AI digest with bilingual output and an MCP interface, and you are comfortable with the fixed source list and the hosted MCP endpoint. Do not adopt it if you need deep customization of sources, want to control the exact schedule beyond the default 07:00 CST, or require a self-hosted MCP without using Cloudflare Workers.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 2 days ago.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 14, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
What agents-radar Solves and Who It Is For
agents-radar solves the problem of manually tracking AI ecosystem news across many scattered platforms. It is for developers, researchers, or product people who want a single daily digest covering GitHub activity, ArXiv papers, Hacker News stories, Hugging Face models, and more. The project is not a news reader you install; it is a pipeline that produces reports automatically. The target user is someone who prefers a curated, bilingual (Chinese and English) summary over browsing each source individually. The README positions it as a daily digest, and the target audience is likely the AI tooling community, given the tracked repos focus on AI CLI tools and agent infrastructure.
The Aggregation Mechanism: Ten Sources and a Fixed Schedule
The core mechanism is a GitHub Actions workflow that runs every morning at 07:00 CST. It pulls data from ten sources, each with a specific API or scraping method. For example, GitHub Repos uses the API to fetch issues, PRs, and releases from 18 tracked AI tool repos. Hacker News uses the Algolia API with six parallel queries to get the top 30 AI stories from the last 24 hours. ArXiv uses its API for papers from cs.AI, cs.CL, and cs.LG in the last 48 hours. Hugging Face uses the Hub API for 30 trending models sorted by weekly likes, but only on Mondays. The schedule is fixed, not configurable per source. The workflow then generates bilingual reports and publishes them as GitHub Issues and committed Markdown files. The Markdown files feed the GitHub Pages site and the RSS feed. The data flow is: source APIs to the workflow, then to Markdown, then to static hosting and feeds.
Getting It Running: Real Commands and Configuration
You do not need to install agents-radar as a package. The README shows setup for the MCP server, which is the only piece you can run locally. To self-host the MCP server, you go to the `mcp/` directory and run `pnpm install` then `wrangler deploy`. That deploys a Cloudflare Worker. For clients, the README gives exact configuration. For Claude Desktop, you edit `~/Library/Application Support/Claude/claude_desktop_config.json` and add an `mcpServers` entry with a `url` pointing to `https://agents-radar-mcp.duanyytop.workers.dev`. For OpenClaw, you run `openclaw mcp add --transport http agents-radar https://agents-radar-mcp.duanyytop.workers.dev` or edit `~/.openclaw/openclaw.json`. The workflow itself is not documented in terms of manual setup; it is assumed to run in the repository's own GitHub Actions. If you fork the repo, you would need to configure the workflow file, but the README does not provide those steps.
The MCP Server: A Convenient Interface with a Hosted Dependency
A notable feature is the hosted MCP server that exposes four tools: `list_reports`, `get_latest`, `get_report`, and `search`. This lets any MCP-compatible client query the latest AI reports conversationally. For example, you can ask Claude "What's the latest in AI CLI tools?" and it calls `get_latest`. The server is hosted at a Cloudflare Workers URL, which means you depend on that endpoint staying up. The README offers self-hosting instructions, but that requires a Cloudflare account and `wrangler`. The MCP server is a differentiator compared to a plain RSS feed, but it is also a single point of failure if the hosted instance goes down. The search tool is limited to recent reports, as the description says "keyword search across recent reports," so it is not a full archive search.
Tracked Repositories: A Narrow but Focused Slice
The project tracks a specific set of GitHub repositories, not the entire AI ecosystem. The README lists AI CLI tools like Claude Code, OpenAI Codex, Gemini CLI, and GitHub Copilot CLI. It also tracks Claude Code Skills from the anthropics/skills repo, with PRs and issues fetched without a date filter and sorted by comment count, so the most discussed skills appear. Additionally, it tracks OpenClaw and peer projects like Hermes Agent, QwenPaw, ZeroClaw, and IronClaw, along with AI infrastructure projects like Ollama, llama.cpp, vLLM, Unsloth, and LiteLLM. This is a curated list, which means the digest will miss AI news outside these repos. The star counts in the README are presentation, not a quality metric. The selection reflects a focus on agent and CLI tooling, not general AI research.
Output Formats: Issues, Markdown, RSS, and a Web UI
The digest is published in multiple formats. GitHub Issues are created daily, and Markdown files are committed to the repo. The Web UI at https://duanyytop.github.io/agents-radar renders those Markdown files via GitHub Pages, with a dark theme and no login. The RSS feed at https://duanyytop.github.io/agents-radar/feed.xml includes the latest 30 reports across all types, updated daily alongside `manifest.json`. This multi-format approach is practical: you can read in a browser, subscribe in an RSS reader, or get notifications via Telegram or Feishu. The README mentions a Telegram channel and a Feishu group, but does not explain how those are populated. The report types are not fully enumerated in the README, but the MCP tools imply there are different types by date and type.
Limitations and Failure Modes
The most obvious limitation is the fixed schedule and fixed source list. You cannot add a source without modifying the workflow code. The Hugging Face update is weekly, on Mondays only, so Tuesday through Sunday digests will miss new trending models. The GitHub Trending component uses HTML scraping, which is fragile if GitHub changes its markup. The ArXiv API only covers three categories, so papers from other AI subfields are excluded. The hosted MCP server is a dependency; if it goes down, the MCP tools fail. The README does not document error handling or retry logic. Also, the last push is unknown and there are no recent releases, so maintenance is uncertain. The project is not archived, but that does not guarantee active development.
Alternatives and How They Differ
A direct alternative is to build a custom GitHub Actions workflow using a tool like `rss-to-issues` or `actions/github-script`, but that requires more effort. Another alternative is using an existing AI news aggregator like `AI News` or `The Rundown`, but those are not open source and do not offer an MCP interface. A more concrete alternative is to use `Hugging Face Daily Papers` which provides a daily digest of ArXiv papers, but it only covers ArXiv and does not generate bilingual reports. The key difference with agents-radar is the integration of multiple sources into one pipeline and the MCP server. If you only need ArXiv papers, Hugging Face Daily Papers is simpler. If you need a self-hosted, customizable aggregator, you would likely write your own workflow. agents-radar is a turnkey solution, but you trade customization for convenience.
Maintenance and License Implications
The project is licensed under MIT, which allows free use, modification, and distribution, with no copyleft obligations. That is permissive for commercial use. The README does not mention contribution guidelines or a changelog. There are no recent releases, and the last push is unknown, so you should check the commit history before relying on it. The MCP server is hosted on Cloudflare Workers, and self-hosting requires a Cloudflare account and `wrangler` deployment. If you fork the repo, you must maintain the workflow yourself. The cost of maintenance is low if the hosted MCP server and GitHub Actions continue to work, but any API changes in the ten sources could break the workflow. The README does not describe update procedures or versioning, so you would need to monitor upstream changes manually.
Editorial conclusion
Adopt agents-radar if you want a zero-maintenance, automated daily AI digest with bilingual output and an MCP interface, and you are comfortable with the fixed source list and the hosted MCP endpoint. Do not adopt it if you need deep customization of sources, want to control the exact schedule beyond the default 07:00 CST, or require a self-hosted MCP without using Cloudflare Workers. Before adopting, verify the GitHub Actions workflow file to confirm the exact configuration keys, the frequency of each source fetch (especially the weekly Hugging Face update), and the current state of the hosted MCP server, since the project has no recent releases and the last push is unknown. Also check the license (MIT) and note that the repository is not archived, but maintenance activity is unclear.
Community notes