Key Features
Agents with Memory
Agents remember past conversations and learn from interactions:
agent = Agent(
model=OpenAIChat(id="gpt-4o"),
storage=PgAgentStorage(table_name="agents", db_url=db_url),
add_history_to_messages=True,
)Knowledge Bases
Add domain knowledge from PDFs, websites, databases, or custom sources:
knowledge = PDFKnowledgeBase(path="docs/", vector_db=PgVector(table="docs"))
agent = Agent(knowledge=knowledge)50+ Built-in Tools
Web search, file I/O, SQL queries, API calls, email, Slack, GitHub, and more.
Agent Teams
Compose specialized agents into teams with a coordinator:
team = Agent(team=[researcher, writer, editor], instructions="...")Structured Output
Get typed responses using Pydantic models — not free text.
Monitoring Dashboard
Production dashboard at phidata.app for logs, metrics, and debugging.
FAQ
Q: What is Phidata? A: A Python framework for building AI agents with memory, knowledge bases, tools, reasoning, and team workflows. Includes production monitoring. 39K+ stars, Apache 2.0.
Q: Is Phidata the same as Agno?
A: Yes, Phidata was rebranded to Agno. The package phidata continues to work and the GitHub repo is the same.