LoginSign Up
← Back to all sparks
A

AutoGen

AI-ASSISTANTS
Velocity0.0

Microsoft open-source framework for building multi-agent AI applications and orchestration.

microsoft.github.io/autogen ↗

AutoGen has gone quiet — last release was September 2025, with no public update for nearly eight months.

release-silenceagent-frameworkmicrosoft-researchpotential-freezeteam-composition
◆Current state
AutoGen's most recent release is python-v0.7.5 on 2025-09-30. The last sustained activity came in a Q3 2025 cluster: v0.7.0 through v0.7.5, with v0.7.1 introducing nested Teams as group-chat participants, RedisMemory, latest MCP version, and OpenAIAgent built-in tools. v0.7.2 made DockerCommandLineCodeExecutor the default for MagenticOne and added an approval_func to CodeExecutorAgent. After that, the cadence stops cold — eight months of public silence as of May 2026.
◆Where it's heading
The technical arc through July–September 2025 was clear: deeper team composition (teams-as-tools, teams-as-participants), better memory (RedisMemory, GraphFlow state retention across resumes), and an MCP-aligned tool surface. Then nothing. For a Microsoft research project in the agent-framework space, an eight-month gap during the most competitive period in agent tooling (LangGraph, OpenAI Agents SDK, Anthropic's Claude Agent SDK, Semantic Kernel agent expansions) is not normal silence — the absence is the signal. Without a release or public roadmap statement, this reads as either pre-major-rewrite mode or quiet wind-down/absorption into another Microsoft framework.
◆Prediction
If there is no release within the next quarter, treat AutoGen as effectively frozen for production use; the agentic framework ecosystem has moved without it. If a release does land, expect it to be a structural rewrite tied to Semantic Kernel or a Microsoft-wide agent surface rather than continuation of the 0.7.x line.

◆Recent moves

  1. 7mo ago

    python-v0.7.5

    python-v0.7.5: thinking-mode support for the Anthropic client, GraphFlow cycle-detection cleanup, Bedrock streaming fixes, RedisMemory linear-memory support. Substantive but routine — and the last public release before the eight-month silence.

    View source ↗
  2. 9mo ago

    python-v0.7.4

    python-v0.7.4: Redis deserialization and streaming-support fixes plus documentation updates. Same-day twin release to v0.7.3 — effectively a maintenance bump.

    View source ↗
  3. 9mo ago

    python-v0.7.3

    python-v0.7.3: adds gpt-5 model info, OpenAIAgent function-tool schema fixes, pydantic extension for anyOf/oneOf, and Redis serialization for complex objects. Useful but incremental.

    View source ↗
  4. 9mo ago

    python-v0.7.2

    python-v0.7.2: makes DockerCommandLineCodeExecutor the default for the MagenticOne team, adds an approval_func option to CodeExecutorAgent (human-in-the-loop hook), and exposes parallel_tool_call on the OpenAI model client. Material defaults change for the multi-agent execution path.

    View source ↗
  5. 9mo ago

    Doc-only patch for OpenAIAssistantAgent

    python-v0.7.1.post1: a single doc-only patch updating the OpenAIAssistantAgent documentation. No code change.

    View source ↗
  6. 9mo ago

    python-v0.7.1

    ⚡ SPARK

    python-v0.7.1 bundles nested Teams as participants in a group chat, RedisMemory, OpenAIAgent built-in tools, an MCP version upgrade, and GraphRAG v2.3. Nested Teams unlocks recursive multi-agent composition — the most structurally significant change in this window, and the last meaningful capability release before the silence.

    View source ↗