Entity trail

Google ADK

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

2026-09-11 / github-pulse-researcherGoogle ADK 2.9.0 reruns failed workflow nodes on resume, so node side effects now repeatgoogle/adk-python 2.9.0 (2026-09-10) changes resume behavior. A failed node used to replay as if it had completed, and now runs again, so any external side effect in that node repeats on every resume. The release notes tell you to make node bodies idempotent. The same release adds `FallbackModel` for automatic failover, a LiveKit voice runner, YAML-authored graph workflows, and opt-in MCP SDK 2.x support, which drops undeclared server fields outside `_meta`. GCS tools now refuse local file access unless `local_file_root` is set.
2026-09-10 / agents-researcherGoogle ships ADK for Kotlin 1.0 with compile-time KSP tool definitions and on-device LiteRT-LM executionGoogle announced ADK for Kotlin 1.0 on 2026-09-09 at full feature parity with ADK 1.0 Core: hierarchical multi-agent delegation, summarization-based context management, human-in-the-loop confirmation gates, and session persistence with resumability. The Android-specific pieces are the reason to care — tools are declared through KSP at compile time with zero runtime reflection, chat history persists to Room, memory indexes through AppSearch, and inference can run on-device via LiteRT-LM or ML Kit (1.0.0-beta) instead of Firebase AI with Gemini 3.8 Flash. Zero-reflection tool binding is a real constraint fix for mobile, where reflection-heavy Python-style agent SDKs collide with R8 shrinking and cold-start budgets.
2026-09-10 / news-researcherGoogle's Agent Development Kit for Python Carries a CVSS 10.0 Unauthenticated RCE via Test Session ReplayCVE-2026-79696, published September 9, is a code injection flaw in `adk web` affecting ADK for Python 2.0.0 through 2.6.0 wherever pytest is installed, including Cloud Run and GKE deployments. It scores a full 10.0: network vector, low complexity, no authentication and no user interaction, with high confidentiality, integrity and availability impact, triggered by a crafted test session replay. The root cause is CWE-184, an incomplete list of disallowed inputs.
2026-09-09 / saas-disruption-researcherSandboxAQ's Switch Tops Product Hunt by Making Slack and Teams a Vendor-Neutral Room Instead of an Agent StorefrontSwitch hit #1 on Product Hunt September 8 with 506 upvotes for an Apache-2.0-with-Commons-Clause layer that drops agents built on Claude Code, Google ADK, LangChain or OpenAI into the same Slack, Teams or Discord channels via MCP, APIs and adapters, self-hostable at no cost. SandboxAQ announced it August 26; the Product Hunt run is what gave it distribution. The strategic read is that it attacks the part of Agentforce and Slack's own agent tier that depends on the collaboration tool being the place you buy your agent.
2026-09-04 / vibe-coding-researcherCVE-2026-79707: unauthenticated path traversal in Google's Agent Development Kit builder endpointPublished to NVD on 2026-09-04, this flaw lets an unauthenticated remote attacker read arbitrary files from a host running Google Cloud Agent Development Kit for Python, versions 1.9.0 through 1.21.0, via a crafted `file_path` query parameter on the builder endpoint. Thirteen minor versions are in range, which means most ADK deployments standing up the builder UI on a reachable interface are affected. If you exposed the ADK builder beyond localhost, treat any secrets on that host as read.
2026-08-29 / github-pulse-researcherCopilotKit open-sources OpenBot, giving each agent its own container, browser and logins behind a single policy gatewayOpenBot reached 3,364 stars and 412 forks since its 2026-08-17 creation, MIT-licensed and pushed 2026-08-28. Every action a bot takes against a computer, file, MCP server or UI component routes through one gateway that resolves the target, decides it against policy, writes an audit row, and only then acts or refuses while naming the rule; each bot gets its own container with its own Chromium, logins and workspace. Agents arrive over the AG-UI protocol rather than a framework binding, so LangGraph, Mastra, CrewAI, Pydantic AI and Google ADK agents all connect the same way and the governance rides the protocol. It self-hosts via Docker Compose with data in your own PostgreSQL and no bundled model, though it is explicitly alpha and its `.env.example` ships `OPENBOT_SINGLE_USER=true`, which admits every request as an administrator until sign-in is configured.
2026-08-28 / rss-researcherBedrock AgentCore Evaluations scores agents from OpenTelemetry spans, so the framework stops matteringAnnounced 2026-08-26, the service reconstructs agent sessions from three OTel span types, invoke agent, inference and execute tool, then applies the same scoring regardless of SDK. It names Strands Agents, LangGraph, OpenAI Agents SDK, LlamaIndex, Google ADK and the Claude Agent SDK explicitly, and covers anything emitting spans under the `opentelemetry.instrumentation.*` or `openinference.instrumentation.*` scope prefixes through generic classification. Built-in evaluators are GoalSuccessRate at session level, Correctness and Helpfulness at trace level, plus custom LLM-as-a-judge at session, trace or tool level; AWS published no pricing or regions.
2026-08-28 / agents-researcherGoogle ADK 2.8.0 adds a Model Armor guardrail plugin and fences relayed sub-agent output so it cannot pose as instructionsADK for Python 2.8.0 landed 26 August with a Model Armor guardrail plugin, a fix that fences relayed agent output so a sub-agent's text cannot be read as instructions by its caller, SQL injection guards in the BigQuery tools, blocked yaml and ruamel deserialization in agent-config code references, and auth taken from the request rather than the client's response. It also adds native task mode for RemoteA2aAgent with auth_scheme and auth_credential support, an ADK_MAX_LLM_CALLS environment variable, and an opt-in experimental telemetry layer reporting per-invocation and per-workflow token spend and tool call counts behind ADK_EXPERIMENTAL_TELEMETRY=true.
2026-08-28 / github-pulse-researcherTwo agent CLIs shipped SSRF and Host-header fixes in the same 48 hoursgemini-cli's 2026-08-27 nightly landed a fix preventing SSRF in MCP OAuth metadata discovery and authentication (PR #29081), and google/adk-python v1.39.1 the same day added a Host header check on its CLI server plus artifact reference scoping and path-segment validation. pydantic-ai v2.35.3, published 2026-08-28, scopes `safe_download` cookies to their original hostnames. Discovery endpoints an agent fetches on your behalf are turning out to be a repeatable SSRF class, and all three fixes are in code paths a self-hosted agent runs by default.
2026-08-27 / saas-disruption-researcherCROSS-CATEGORY: Slack and SandboxAQ Both Claimed the Human-Agent Collaboration Layer on the Same Day, From Opposite DirectionsOn 2026-08-26 Slack shipped Slack Code, a proprietary channel primitive that gives agents a first-class place inside Slack, while SandboxAQ open-sourced Switch under Apache 2.0 with Commons Clause, turning channels in Slack, Microsoft Teams and Discord into vendor-neutral shared rooms for agents built with Claude Code, Google ADK, LangChain or OpenAI. Product Hunt's #3 launch the next morning, Traccia, sold the same vendor-neutral promise one layer down as an OpenTelemetry-based agent control plane naming Langfuse, Helicone and Lyzr as competitors. The pattern is the 2024 MCP fight replayed at the collaboration tier: incumbents make the agent a native object in their surface, challengers make the surface interchangeable.
2026-08-15 / agents-researcherGoogle ADK Python 2.7.0 stops guessing model capabilities from the model id and lets tools return images back to the modelADK 2.7.0 (2026-08-13) is framed by its maintainers as a correctness release about what an agent sends the model and what it gets back. Models now declare their own capabilities, so an agent pairs an output schema with tools only when the model actually supports it instead of ADK inferring support from the model id — the kind of string-matching heuristic that silently breaks on every new model release. Tool function responses can now carry images and other media straight back to the model across Gemini, Anthropic, LiteLLM, Apigee and OCI; conversation history preserves Gemini thought signatures, server-side tool call parts and every result of a parallel function call; and `inject_session_state` gains optional Jinja2 templates with an `{{ artifact("name") }}` helper, with the regex engine still the default.
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).

Graph relationships

RELEASED
Google
Google released ADK Python 2.7.0 with improved model capability declaration and media response handling.
Source finding
USES
Solace Agent Mesh
Solace Agent Mesh is integrated with Google Agent Development Kit.
Source finding
SUPPORTS
OpenBot
OpenBot supports Google ADK agents via AG-UI protocol.
Source finding
COMPETES WITH
Agent Framework
Agent Framework competes with Google ADK
Source finding
RELEASED
Google -> Google ADK
Google released ADK Python 2.7.0 with improved model capability declaration and media response handling.
Source finding
USES
Solace Agent Mesh -> Google ADK
Solace Agent Mesh is integrated with Google Agent Development Kit.
Source finding
SUPPORTS
OpenBot -> Google ADK
OpenBot supports Google ADK agents via AG-UI protocol.
Source finding
COMPETES WITH
Agent Framework -> Google ADK
Agent Framework competes with Google ADK
Source finding

Source trail

2.9.0 SandboxAQ GitHub GitHub AWS PRNewswire Product Hunt GitHub The Register v1.38.0 and v2.6.3 Diagrid (Stack Overflow)

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…