OpenClaw -- the open-source AI assistant formerly known as ClawdBot -- has taken the developer world by storm. With over 247,000 GitHub stars and a skills registry that has surpassed 13,000 community-built plugins, it has become the default way for developers and power users to run a personal AI assistant on their own hardware.
But out of the box, OpenClaw has a blind spot: it does not know what happened today. Like any LLM-powered assistant, it is limited by its training data cutoff. Ask it about a breaking story, a recent earnings report, or yesterday's policy announcement, and it will either admit it does not know or hallucinate something plausible.
A news API skill fixes this. In this guide, we will build a custom OpenClaw skill that connects to the NewsMesh API, giving your assistant real-time access to news from thousands of sources worldwide. By the end, your OpenClaw instance will be able to answer questions like "What happened with NVIDIA today?" or "Give me the latest technology news" with accurate, sourced information -- all from within WhatsApp, Telegram, Slack, or whatever channel you use.
What Is OpenClaw?
OpenClaw is a personal AI assistant you run locally on your Mac, Windows, or Linux machine. Created by Peter Steinberger, it was originally called ClawdBot before being renamed to Moltbot and finally OpenClaw.
The core concept is simple: instead of using a web-based chatbot, you run an assistant on your own hardware that connects to the messaging apps you already use. You message it on WhatsApp or Telegram, and it responds by routing your request through an LLM (Claude, GPT, DeepSeek, or a local model) combined with skills that give it the ability to take real actions.
What makes OpenClaw different from other AI assistants:
- Local-first architecture. Your data stays on your machine. The Gateway control plane runs at
ws://127.0.0.1:18789and binds to loopback by default. - Multi-channel access. Talk to your assistant through WhatsApp, Telegram, Discord, Slack, Signal, iMessage, and 15+ other platforms.
- Extensible via skills. Skills are modular plugins defined in simple markdown files that teach the assistant new capabilities.
- Persistent memory. The assistant remembers your preferences and maintains context across conversations.
- Self-improving. It can create and modify its own skills autonomously when you ask it to handle a new task.
The skills system is where things get interesting for developers. Instead of writing complex integrations, you describe what the skill does in a SKILL.md file, specify the API endpoints and parameters, and OpenClaw's agent figures out when and how to use it. This makes it trivially easy to connect any REST API -- including a news API.
Why Your AI Assistant Needs Real-Time News
An AI assistant that cannot access current information is limited to being a smart search engine for static knowledge. Adding real-time news access transforms it into something genuinely useful for daily workflows:
Morning briefings. Message your assistant "Give me a 5-minute briefing on tech and business news" and get a concise summary of what matters today.
Research and monitoring. Track topics, companies, or industries. "What's the latest news about Apple?" returns sourced articles from the past few hours, not training data from months ago.
Event awareness. Your assistant can proactively understand the context of your questions. When you ask about a company's stock price, it can correlate with recent news coverage.
Multi-source perspective. A news API pulls from thousands of publishers simultaneously, giving you broader coverage than any single outlet.
The NewsMesh API is well-suited for this because it returns structured data -- title, description, source, category, topics, and publication timestamp -- that maps cleanly to how an LLM agent needs to consume information. It is not raw HTML scraping. It is clean, queryable data ready for an AI to work with.
Building the NewsMesh Skill for OpenClaw
OpenClaw skills are defined in a SKILL.md file placed in your workspace's skills directory. The file contains YAML frontmatter with metadata, followed by natural-language instructions that tell the agent how to use the skill.
Step 1: Get a NewsMesh API Key
Sign up at newsmesh.co/signup and grab your API key from the dashboard. The free tier gives you 25 requests per day, which covers basic personal use.
Step 2: Configure the API Key
Never embed API keys inside skill files. Instead, add the key to your OpenClaw configuration at ~/.openclaw/openclaw.json:
{
"skills": {
"entries": {
"newsmesh": {
"env": {
"NEWSMESH_API_KEY": "your-api-key-here"
}
}
}
}
}
This injects the NEWSMESH_API_KEY environment variable into the agent's process when the skill runs.
Step 3: Create the Skill File
Create the directory and file at ~/.openclaw/skills/newsmesh/SKILL.md:
---
name: newsmesh
description: Search and retrieve real-time news articles from thousands of sources worldwide using the NewsMesh API. Supports searching by keyword, browsing latest headlines, filtering by category and country, and getting trending stories.
requires:
env:
- NEWSMESH_API_KEY
bins:
- curl
- jq
---
# NewsMesh News API Skill
You have access to the NewsMesh news API for retrieving real-time news articles.
## Base Configuration
- Base URL: `https://api.newsmesh.co/v1`
- Authentication: Pass the API key in the `X-Api-Key` header
- All responses return JSON with a `data` array and a `next_cursor` field for pagination
## Available Endpoints
### Search Articles
Search for news articles matching a query.
```bash
curl -s "https://api.newsmesh.co/v1/search?q=QUERY&limit=LIMIT" \
-H "X-Api-Key: $NEWSMESH_API_KEY" | jq
Parameters:
q(required): Search query stringlimit(optional): Number of results, default 25
Latest Articles
Get the most recent articles, optionally filtered.
curl -s "https://api.newsmesh.co/v1/latest?limit=LIMIT&category=CATEGORY&country=COUNTRY" \
-H "X-Api-Key: $NEWSMESH_API_KEY" | jq
Parameters:
limit(optional): Number of results, default 25category(optional): Filter by category. Values: business, technology, science, health, sports, entertainment, politics, world, environment, lifestylecountry(optional): Filter by 2-letter ISO country code (e.g., us, gb, de, in)from(optional): Start date in YYYY-MM-DD formatto(optional): End date in YYYY-MM-DD format
Trending Articles
Get currently trending news articles.
curl -s "https://api.newsmesh.co/v1/trending?limit=LIMIT" \
-H "X-Api-Key: $NEWSMESH_API_KEY" | jq
Parameters:
limit(optional): Number of results, default 25
Response Format
All endpoints return {"data": [...], "next_cursor": "..."}. Each article in the data array contains:
article_id: Unique article identifiertitle: Article headlinedescription: Brief summarylink: URL to the full articlesource: Publisher namepublished_date: Publication timestamp (ISO 8601)category: Article categorymedia_url: Thumbnail image URL (if available)topics: Array of topic tagspeople: Array of people mentioned (if any)author: Article author (if available)
Guidelines
- When the user asks about current events, recent news, or "what happened with X", use the search endpoint.
- When the user asks for a news briefing or latest headlines, use the latest endpoint.
- When the user asks what's trending, use the trending endpoint.
- Always include the source name and a link when presenting articles.
- Summarize articles concisely -- present the key facts, not the full text.
- When multiple articles cover the same story, synthesize them into a single summary and note the sources.
- Default to 5 articles unless the user asks for more.
- If the user asks about a specific category (e.g., "tech news"), use the category filter.
- If the user asks about news in a specific country, use the country filter.
This skill file is everything OpenClaw needs. When you ask your assistant about news, it will recognize the intent, load this skill, and execute the appropriate `curl` command to fetch articles from the NewsMesh API. It then uses the LLM to summarize and present the results conversationally.
### Step 4: Test the Skill
Verify the skill is loaded correctly:
```bash
claw skill test newsmesh
Then message your assistant through any connected channel:
"What's the latest technology news?"
OpenClaw will load the NewsMesh skill, call the /v1/latest endpoint with category=Technology, and present you with a summarized briefing of recent tech headlines.
Real-World Usage Patterns
Once the skill is installed, here are the patterns that work well in day-to-day use.
Daily News Briefings
Combine the skill with OpenClaw's cron scheduling to get automated briefings:
Set up a daily briefing: every morning at 8am, get me the top 5 trending news stories and the top 3 stories in Technology and Business.
OpenClaw can create a scheduled task that calls the NewsMesh skill on a cron schedule and sends the results to your preferred channel.
Topic Monitoring
Track specific companies, people, or topics over time:
"Search for news about OpenAI in the last 24 hours."
"What's happening with the EU AI Act this week?"
"Any news about SpaceX launches?"
The search endpoint handles natural-language queries effectively because NewsMesh uses full-text search under the hood, matching against article titles and descriptions.
Research and Context Gathering
When you are working on something and need background context:
"I'm writing about renewable energy trends. Get me the latest 10 articles about solar energy and wind power."
"Find recent news about the Series B funding round for Mistral AI."
News-Aware Conversations
The skill works alongside OpenClaw's persistent memory. If you have been discussing a topic with your assistant, it can pull in relevant news to enrich the conversation:
"We were talking about the semiconductor shortage yesterday. Any new developments?"
The assistant remembers the prior conversation context and uses it to construct a targeted news search.
Advanced: Combining with Other Skills
OpenClaw skills compose naturally. Here are patterns that combine news data with other capabilities.
News + Summarization
If you have a long-form content skill or document writing skill installed, you can chain them:
"Get the top 10 articles about climate change this week and write me a one-page executive summary."
The assistant fetches articles via the NewsMesh skill, then uses its writing capability to produce a structured summary.
News + Calendar
With a calendar skill installed:
"Check if there's any major tech news today that's relevant to my meetings tomorrow."
News + Notification Filtering
You can ask OpenClaw to set up smart notifications:
"Monitor news about my portfolio companies (Tesla, Apple, NVIDIA) and only alert me if something significant happens -- like earnings reports, lawsuits, or executive changes."
OpenClaw can periodically query the NewsMesh API via a cron job, evaluate the results using the LLM, and only forward articles that meet your criteria.
Performance and Cost Considerations
A few practical notes for running this in production.
API limits. The NewsMesh free tier provides 25 requests per day (750 per month). That covers a morning briefing plus a handful of ad-hoc queries. If you plan to use automated cron jobs or make more than a few queries per day, the Starter plan at $29/month ($23/month annual) provides 5,000 requests per month.
Response time. NewsMesh API responses are fast. Combined with OpenClaw's LLM processing time, expect a few seconds end-to-end from sending a message to receiving a formatted news summary. This is fast enough for conversational use.
Caching. For skills that make repeated similar queries (like a morning briefing), OpenClaw's session system naturally avoids redundant calls within the same conversation. For cross-session caching, you can extend the skill to write results to a local file and check freshness before making a new API call.
Token usage. Each article returned by the API is roughly 100-200 tokens (title + description + metadata). Fetching 10 articles adds about 1,500 tokens of context to the LLM prompt. This is efficient -- far less than scraping full web pages and much more structured.
Security Considerations
Since OpenClaw runs locally and skills can execute shell commands, security matters.
API key handling. As covered above, keep keys in openclaw.json configuration, not in the skill file itself. If your skill file is shared publicly (for example, on ClawHub), it must not contain credentials.
Input sanitization. The skill file uses $NEWSMESH_API_KEY as an environment variable in curl commands. OpenClaw handles the variable substitution, but if you extend the skill with custom scripts, validate any user input that gets interpolated into shell commands.
Rate limiting. The NewsMesh API enforces rate limits server-side, so even if a runaway cron job makes excessive requests, your account will not incur unexpected charges -- requests beyond your plan's limit simply return a 429 status.
DM pairing. Make sure your OpenClaw instance has DM pairing enabled (the default) so that only approved users can trigger skills. An open DM policy on a publicly accessible channel could let strangers consume your API quota.
Conclusion
OpenClaw's skill system makes it remarkably easy to give your AI assistant capabilities that used to require building a full application. A single markdown file turns your personal assistant into a news-aware agent that can search, summarize, and monitor real-time news from thousands of sources.
The NewsMesh API is a natural fit for this pattern. It provides structured, queryable news data through simple REST endpoints -- exactly what an OpenClaw skill needs. No web scraping, no HTML parsing, no authentication complexity. Just clean data delivered fast.
To get started, sign up for a free NewsMesh API key, create the skill file shown above, and start asking your assistant about the news. For more on what you can build with news API data, see our guide on using news APIs to power AI agents and RAG pipelines and our complete guide to news APIs.