python-1.10.0

Published June 30, 2026 · Category: AI Agents

Overview

[1.10.0] - 2026-06-30

Added

• agent-framework-core: Explicitly emit available_resources and available_scripts in skill content (#6694)
• agent-framework-core: Autolabelling MCP servers based on hints and GitHub MCP server interface labels (#6171)
• agent-framework-core: create_harness_agent skills_paths accepts str | Path | Sequence[str | Path] | None (#6717)
• agent-framework-core: Background agent loop resolves provider automatically and adds feedback message builder (#6735)
• agent-framework-core: Stop swallowing skill script and resource errors so the model can self-correct (#6755)
• agent-framework-foundry: Add FoundryAgent conversation session helper (#6623)
• agent-framework-foundry: Add support for Foundry Adaptive evals (#6267)
• agent-framework-bedrock, agent-framework-gemini: Surface cache and reasoning token counts (#6640)
• agent-framework-durabletask: Host MAF workflows on a standalone Durable Task worker (#6418)
• agent-framework-azure-ai-search: Support stable and preview Azure AI Search (Foundry IQ) API versions (#6603)
• agent-framework-github-copilot: Align function approval to use SDK on_pre_tool_use hook (#6750)

Changed

• agent-framework-core: [BREAKING — experimental] Refactor FileSkillsSource for depth-based discovery and predicate filters (#6488)
• agent-framework-core: [BREAKING — experimental] Require approval for file-access tools with read-only auto-approval (#6599)
• agent-framework-core: [BREAKING — experimental] Integrate looping into HarnessAgent (#6607)
• agent-framework-core: [BREAKING — experimental] Port FileMemoryProvider and integrate into harness agent (#6547)
• agent-framework-core: [BREAKING — experimental] Make all SkillsProvider tools require approval by default (#6754)
• agent-framework-core: [BREAKING — experimental] Improve FileAccess/FileMemory harness providers (surgical edits, read-only tier, consistent naming) (#6801)
• agent-framework-core: Align serialized tool format to OTel GenAI tool definition format (#6556)
• agent-framework-foundry-hosting: [BREAKING] Foundry Hosted Agent V2 protocol upgrade (#6811)
• agent-framework-foundry-hosting: Add MCP as a hard dependency (#6634)
• agent-framework-purview: Prefer token principal for user identity (#6693)
• agent-framework-ollama: Convert Pydantic model class response_format to JSON schema in OllamaChatClient (#6782)

Fixed

• agent-framework-core: Fix background agent telemetry context error (#6764)
• agent-framework-core: Fix MCP metadata and tool name handling (#6656)
• agent-framework-core: Ensure spans created inside sync preparations in streaming call are correctly nested (#6552)
• agent-framework-openai: Preserve OTel parent context for deferred streams (#6709)
• agent-framework-openai: Fix FunctionShellTool throw and empty streaming shell command (#6763)
• agent-framework-ag-ui: Fix tool history replay sanitization (#6581)
• agent-framework-anthropic: Re-role trailing assistant message to user for Anthropic compatibility (#6207)
• agent-framework-hyperlight: Fix CodeAct span parenting (#6712)
• agent-framework-hyperlight: Harden output capture against symlinks (#6601)

Source

Originally published at github.com.

Related Articles

• NVIDIA BioNeMo Agent Toolkit Brings Accelerated AI to Life Sciences Researchers in Claude Science
• How NVIDIA’s Inference Software Stack Powers the Lowest Token Cost
• How Jaiveer Singh Is Helping Robots — and Developers — Move Faster