Investment Committee
Explore a standalone investment-research team application.
This standalone walkthrough follows investment-team at the reviewed revision, whose Docker requirements pin Agno 2.5.2.
Investment Committee is a separate standalone application, not a registered team in the reviewed Demo OS. The linked repository installs its pinned requirements through Docker and runs locally on port 8000.
Why a team for this
Specialists contribute market, financial, technical, risk, knowledge-retrieval, and memo-writing work. Role prompts organize perspectives; they do not eliminate bias or establish the reliability of a recommendation.
Four team modes, one set of agents
The application registers seven agents, including the Committee Chair, four teams, and an investment workflow. Select the actual team ID in its own runtime:
| Team ID | Pattern |
|---|---|
| coordinate-team | The leader coordinates specialist work and synthesizes results |
| route-team | Dispatch to a selected member |
| broadcast-team | Gather member responses to the same task |
| task-team | Decompose and manage work as tasks |
Three-layer knowledge
Research knowledge, reusable lessons, and memo files are separate resources. The knowledge agent has research knowledge; the memo writer has file tools. They do not all share an identical Knowledge and LearningMachine configuration.
Shared learnings
The chair and participating analysts use learned knowledge according to their own configuration. The shared source names are team_knowledge and team_learnings. Inspect which agents receive each resource before assuming universal access.
File output
The memo writer saves files under the root memos/ directory. Its naming instructions use ticker, year, quarter, and recommendation. Inspect the generated file and supporting evidence.
Mixing models
The reviewed source configures gemini-3-flash-preview for specialists and gemini-3.1-pro-preview for chairs/team models. These are the application’s selected model IDs, not evidence of a measured quality advantage.
The explicit workflow runs market assessment, parallel financial and technical analysis, risk assessment, memo writing, and a final committee decision.
See it in action
Use a separate local checkout and sample data:
git clone https://github.com/agno-agi/investment-team.git
cd investment-team
cp example.env .env
# Set GOOGLE_API_KEY and EXA_API_KEY. PARALLEL_API_KEY is optional.
docker compose up -d --build
docker exec -it investment-team-api python -m app.load_knowledgeConnect http://localhost:8000 in the AgentOS UI and choose one of the team IDs above. The default application has no production authorization configuration; this is a local demonstration.