Serve and embed

Put the research team and pipeline behind an API so anyone, and any schedule, can ask for a review.

A research system that only runs from a script helps the person at the keyboard. Behind an API it answers from Slack, runs as a nightly review, and feeds a dashboard. AgentOS serves agents, teams, and workflows on the same contract.

from agno.os import AgentOS

agent_os = AgentOS(
    agents=[market_analyst, financial_analyst, risk_officer, committee_chair],
    teams=[coordinate_team, broadcast_team],
    workflows=[investment_workflow],
)
app = agent_os.get_app()

if __name__ == "__main__":
    agent_os.serve(app="main:app", port=8000)

Every registered agent, team, and workflow gets run and session endpoints. A surface calls the workflow for the auditable review or a team for an open-ended question, with the same request shape.

curl -X POST http://localhost:8000/workflows/investment-workflow/runs \
  -F 'message=Run a full review on NVDA' \
  -F 'user_id=[e-mail]' \
  -F 'session_id=q4-review' \
  -F 'stream=false'

Where a research system lives

Surface	Shape
Slack channel	An analyst asks “review NVDA”; the pipeline replies in-thread with the memo
Scheduled review	A cron runs the workflow nightly on a watchlist and posts the decisions
Dashboard	A widget triggers a review and renders the structured decision
Backend gate	A pipeline calls the workflow before a position changes, blocks on a PASS

The serving model is identical to a product agent and a data agent. The difference is what runs behind the endpoint, not how it is served.

Pick the entry point per request

Caller wants	Hit
The decision of record	The Workflow (deterministic, auditable)
An open-ended question	A coordinate Team
An independent multi-view read	A broadcast Team
A single fact	A route Team

One AgentOS exposes all of them. The surface chooses the shape per request.

Scheduled research is the high-leverage pattern

The review nobody remembered to run is the one that mattered. A scheduled workflow turns deep research from on-demand to always-on: every morning the watchlist is reviewed, the memos are written, and the decisions are waiting before anyone asks.

Next steps

Task	Guide
Add Slack or a browser surface	Interfaces
Lock down the endpoints	Security and auth

Developer Resources

• Serve as an API
• AgentOS cookbook
• Scheduling