OpenAI Agents SDK for Python v0.23.0
sdkOpenAI Agents SDK for Python v0.23.0 adds configurable MCP listing page limits, opt-in Docker sandbox removal protection, configurable sandbox memory consolidation turns, and an opt-in encrypted session history scan budget. The release also ships numerous core fixes around nested agent-tool state isolation, tool identity preservation, guardrail diagnostics persistence, and redaction of streaming/tool failure tracebacks.
An agent can now page MCP tool listings with configurable limits, protect its Docker sandbox from accidental removal, tune memory consolidation turns, and bound encrypted session history scans.
Minor feature release: four new opt-in features (MCP paging, Docker removal protection, memory consolidation turns, encrypted history scan budget) plus a large batch of core correctness fixes around nested agent tools and redaction.
pip install openai-agents==0.23.0from agents import Agent, Runner
from agents.mcp import MCPServerStdio
# New in 0.23.0: configurable MCP list_tools page limit
mcp_server = MCPServerStdio(
params={"command": "uvx", "args": ["mcp-server-fetch"]},
list_tools_page_limit=50,
)
agent = Agent(
name="researcher",
instructions="Use MCP tools to answer.",
mcp_servers=[mcp_server],
)
result = Runner.run_sync(agent, "Fetch https://example.com and summarize.")
print(result.final_output)