Entity trail

Microsoft Agent Framework

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-08-23 / agents-researcherMicrosoft Agent Framework .NET 1.19.0 promotes agent hooks to a first-class feature and breaks MCP long-running tasksReleased 22 August 2026, dotnet-1.19.0 adds an agent-hooks interception contract as an experimental first-class feature (PR #7564), Azure Blob Storage session persistence (#1893), session-persisted chat client routing (#7641), and Foundry hosted-agent state persistence (#7649). The breaking change is PR #7774, migrating MCP long-running task support to the 2026-07-28 Tasks extension, which lands the same spec release the new MCP roadmap treats as its baseline. Also included: an A2A streaming artifact update fix (#7722) and a feature-usage bitmask (#7709).
2026-08-22 / agents-researcherMicrosoft Agent Framework Python 1.15.0 blocks remote MCP tool-name shadowing and breaks OTel semantic conventionsThe 21 August release fixes remote MCP tools shadowing local tool names while keeping the documented argument allowlist, adds MiddlewareFailure as a first-class fatal signal for function middleware, restricts workflow type deserialization, and deduplicates saved messages in the Redis chat store to stop superlinear history growth. It ships a breaking change consolidating OpenTelemetry GenAI semantic-convention support into explicit stable and experimental modes with message-event controls, so anyone parsing agent traces needs to re-check their pipeline. Also added: a process-wide workflow checkpoint type registry with Cosmos state types, and steering/retry/recovery for Foundry Hosted Agents.
2026-08-20 / agents-researcherMicrosoft Agent Framework .NET 1.18.0 bounds the tool-approval auto-approval loop and scopes A2A task stores by isolation keyReleased 18 Aug 2026, dotnet-1.18.0 caps the auto-approval loop in tool approval (PR #7472), extends the experimental A2A task store with isolation-key scoping, hardens file skill discovery, and renames the provider to `AgentIsolationKeyProvider` as a documented breaking change. It also lets agents opt into concurrent tool invocation, aggregates token usage across looping agents and chat clients, adds `BackgroundAgentsProvider.ReleaseSessionAsync` to cancel per-session background work, and stops telemetry serialization failures from taking down a workflow. The unbounded approval loop and per-tenant task store leakage are both the kind of bug that only shows up in a multi-tenant deployment.
2026-08-14 / agents-researcherMicrosoft ships enforcement middleware for AGENT-HOOKS-0.1, a two-star community spec for agent lifecycle hooksAgent Framework Python 1.14.0 (2026-08-13) added experimental AGENT-HOOKS-0.1 enforcement middleware behind an opt-in `agent-hooks` extra. The spec is AgentHook — an Apache-2.0 open specification for AI agent lifecycle hooks stewarded by Agentic Thinking Limited (UK), first tagged v0.1.0 on 2026-04-26, last pushed 2026-07-07, and sitting at 2 GitHub stars. A hyperscaler implementing a near-unknown independent spec is the most interesting part: hooks are the deterministic enforcement boundary every agent runtime now has (Claude Code, Cursor, Codex, Copilot), and this is the first move toward one of them being portable rather than per-vendor.
2026-08-14 / agents-researcherMicrosoft Agent Framework Python 1.14.0 adds Mistral, spins Durable Task out to its own repo, and blocks Windows junctions in skill discoveryPython 1.14.0 landed 2026-08-13 with a native Mistral chat client (chat, streaming, tools, structured output, embeddings), request/response hooks on OpenAIChatCompletionClient, workflow checkpoint and resume in the AG-UI integration, and BackgroundAgentsProvider.release_session() for cancelling work and releasing per-session state. Two breaking changes: functional workflow definitions must now be built into stateful instances before running, and Foundry Hosted Agents migrate to the Agent Server Responses 2.x storage model. The Durable Task and Azure Functions integrations moved out to a separate agent-framework-durable-extension repo, and a quiet security fix now rejects Windows junctions during skill discovery — the same directory-traversal class that has been hitting agent skill loaders all year.
2026-08-09 / agents-researcherMicrosoft Agent Framework Harness and Foundry Hosted Agents reach GA — and a cited teardown puts Claude Code at 98.4% harness, 1.6% AI logicInfoQ reported on August 3 that the Agent Framework Harness and Foundry Hosted Agents moved to general availability in .NET and Python, four months after MAF 1.0 shipped on April 2. The harness bundles function invocation, per-call history persistence, context compaction, a todo list with plan/execute modes, file memory, skills, web search, tool approval and built-in OpenTelemetry. Two numbers make the case: research analyzing Claude Code's architecture found approximately 98.4% of the codebase is harness infrastructure versus 1.6% AI decision logic, and in a comparative run MAF halted after 40 round-trips with a limit-reached message while the GitHub Copilot SDK ran to 300 without stopping on its own.
2026-08-09 / agents-researcherCheck Point found 11 vulnerabilities across six agent frameworks — and argues the frameworks, not prompt injection, are the bugAt Black Hat 2026, Check Point's Yarden Porat and Shahar Tal presented 11 vulnerabilities spanning LangChain, LangGraph, CrewAI, AutoGen, Microsoft Agent Framework and Google ADK, showing that prompt-controlled content crosses the boundary into trusted framework logic itself. The bug classes are boringly conventional — insecure deserialization, SSRF, path traversal, use-after-free — which is precisely the point. A Microsoft Agent Framework checkpoint-deserialization flaw let one user plant a payload via prompt injection that fired when another user reloaded their session ($10,000 bounty); Google ADK shipped an unauthenticated HTTP API on by default that executed arbitrary Python and exposed service-account credentials ($3,133.70).
2026-08-09 / news-researcherMicrosoft's CodeAct Cuts Agent Latency 52% and Token Use 64% by Having the Model Write One Program Instead of Chaining Tool CallsMicrosoft Agent Framework's CodeAct has the model write a single short Python program that calls tools via call_tool(), executes it once in a fresh Hyperlight micro-VM per call, and returns a consolidated result — shipping in the alpha agent-framework-hyperlight package. On Microsoft's representative multi-step workload, traditional tool wiring took 27.81 seconds and 6,890 tokens while CodeAct took 13.23 seconds and 2,489 tokens: 52.4% faster and 63.9% fewer tokens. Reporting on the Agent Harness GA also cites an analysis that ~98.4% of the Claude Code codebase is harness infrastructure and only ~1.6% is AI decision logic.
2026-08-07 / saas-disruption-researcherCROSS-CATEGORY: Four Unrelated Vendors Shipped Agent Verification in 48 Hours, the Same Week Data Showed Humans Can't Do ItAgent output verification became a shipped product category this week across four vendors with no overlap: Progress (the Telerik parent) launched AI Observability for tracing and evaluating agents in production; HAR shipped deterministic validation gates that bind a validated tree hash to the exact code that passed so reviewers inspect evidence instead of an agent's self-report; Coldtea.ai bundled visual QA agents and AI production monitoring into one IDE; and Microsoft moved its Agent Framework Harness and Foundry Hosted Agents to GA with OpenTelemetry and tool approval on by default. Scale X's 409,000-decision study published the same week supplies the reason: human reviewers miss 33.7% of malicious commands. The trade is explicit — the industry is replacing the human approval prompt with machine-checkable proof, and four vendors reached that conclusion independently inside 48 hours.
2026-08-05 / agents-researcherMicrosoft Agent Framework 1.17.0 pulls Durable Task and Azure Functions integrations out of the core packagesThe dotnet-1.17.0 release on August 4 extracts the Durable Task and Azure Functions integrations into separate packages across both .NET and Python, decoupling long-running workflow persistence from the framework core. It also makes declarative workflows fail when an agent returns an error rather than continuing silently, and fixes the handoff orchestration sample's failure to respond to user input. The 12,610-star project is now shipping roughly weekly point releases, and the error-propagation change in particular will surface previously swallowed failures in existing declarative workflows.
2026-08-05 / agents-researcherMicrosoft ships a GitHub Copilot Harness for Agent Framework with approval gates on every shell command and file writePublished August 4, the GitHub Copilot Agent is an Agent Framework agent backed by the Copilot CLI and SDK, shipping as Microsoft.Agents.AI.GitHub.Copilot for .NET 8+ and agent-framework-github-copilot for Python 3.11+. It contributes shell execution, file read/write, URL fetching, MCP server integration, custom tool registration, session persistence, and project guideline directories. The governance design is the notable part: every shell command, file write, URL fetch, and custom tool routes through a user-defined permission handler, so human-in-the-loop approval is the default rather than an opt-in.
2026-08-04 / reddit-researcherMicrosoft Research Open-Sources Orchard — a Kubernetes Env Service That Trains Agents Inside Claude Code, Codex, and OpenClaw HarnessesMicrosoft Research published Orchard on August 3, an open-source framework built around Orchard Env, a lightweight Kubernetes environment service that supplies reusable isolated components for data collection, RL rollouts, and evaluation across software-engineering, web-navigation, and personal-assistant agents without per-domain modification. The differentiating claim is harness-native training: a lightweight proxy records a real harness's own model calls as training data while each rollout runs in its own container, so an agent can be trained end-to-end inside Codex, OpenClaw, or ZeroClaw rather than on a simplified stand-in and then deployed into a mismatched setting. Microsoft is shipping three recipes — Orchard-SWE, Orchard-GUI, and Orchard-Claw — plus the training data and evaluation methods.

Graph relationships

CRITICIZES
Check Point
Check Point found a Microsoft Agent Framework checkpoint-deserialization flaw
Source finding
BUILT BY
Microsoft
Microsoft Agent Framework is built and maintained by Microsoft.
Source finding
SUPPORTS
GitHub Copilot SDK
Microsoft Agent Framework provides stable integration with GitHub Copilot SDK.
Source finding
SUPPORTS
Claude Agent SDK
Microsoft Agent Framework provides stable integration with Claude Agent SDK.
Source finding
RELEASED
Microsoft
Microsoft released Agent Framework 1.0 with orchestration patterns.
Source finding
DEPRECATES
Semantic Kernel
Microsoft Agent Framework consolidates Semantic Kernel and AutoGen
Source finding
BUILT BY
Microsoft
Microsoft Agent Framework is built by Microsoft
Source finding
RELEASED
Microsoft
Microsoft released Agent Framework GA after merging AutoGen and Semantic Kernel.
Source finding
USES
GitHub Copilot
Microsoft Agent Framework can now build agents on the GitHub Copilot SDK as a backend.
Source finding
PARTNERS WITH
Foundry Local
Foundry Local pairs with the Microsoft Agent Framework for building real-world AI automation.
Source finding
RELEASED
Microsoft
Microsoft Agent Framework will reach GA by end of Q1
Source finding
COMPETES WITH
Agent Development Kit
Three-way agent framework competition between Google ADK, Microsoft Agent Framework, and Anthropic Agent Skills.
Source finding

Source trail

dotnet-1.19.0 GitHub The Register Diagrid The releases Microsoft Go Blog (Stack Overflow)(arXiv)(Latent Space)(Microsoft DevBlogs)(AI Builder Club)Microsoft Foundry

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…