Build your first agent
Use your coding agent to build your first agent
After setting up your platform, your coding agent will ask what you want to build. You could create an agent to triage issues, answer questions in Slack, query databases, or review invoices and contracts.
I recommend starting with a Daily Brief, an agent that researches the companies, products, and topics you follow. Each morning, it tells you what changed, explains why it matters to you, and links to the sources. You can ask follow-up questions and change what it covers.
We'll start with a briefing about agent platforms and AI tools. Replace those interests with your own:
| Your role | What you might follow |
|---|---|
| Marketing | Competitor launches, positioning, customer concerns, and ideas for useful content. |
| Sales | Target-account announcements, leadership changes, partnerships, and competitor updates. |
| CTO | Model and infrastructure releases, pricing changes, security advisories, and deprecations. |
| Engineering leader | Releases affecting your stack, migration deadlines, and practical engineering lessons. |
Create your first agent
Use the prompt below to create your first agent. If you're starting a new conversation, add /create-agent so your agent knows to use the skill.
Build an agent called Daily Brief.
I work on AI developer tools. Follow agent frameworks and platforms,
model and coding-agent releases, and practical examples of agents
running in production. Prioritize developments that could affect
what we build, how we explain it, or who might use it.
Skip generic AI commentary.
Use the Parallel MCP to research the web.
Produce a three-piece edition with one lead story and up to two
other developments. For each, explain what changed, why it matters
to me, and link to dated original sources. Prefer developments since
the previous edition; use the last 24 hours for the first edition.
Save previous editions in the shared notes.
Use per-user learning to remember each user's preferences.Change the interests and other details before sending. Parallel's MCP search and fetch tools offer free access at lower rate limits without an API key. Model calls still use your configured model credentials.
Your coding agent will create an agent in agents/, registers it in app/main.py, and adds a description and sample prompts to app/config.yaml.
Test your first agent
Refresh the Agents list in AgentOS, open Daily Brief, and ask it to generate today's edition. Try a follow-up:
- "Explain the lead story in more detail."
- "More implementation examples, less funding news. Remember that for future editions."
- "What has changed since the previous edition?"
Inspect the session and trace to see the searches and page reads behind the edition. If something is wrong, give your coding agent the example and ask it to fix the behavior.
Next
Deploy your platform to Render, secure it, and enable the daily brief on your live connection.