{"company":{"name":"Harness","slug":"harness","website":"https://harness.io","category":"developer-tools"},"question":"What has Harness shipped recently?","answer":"In the last 30 days, Harness shipped 54 tracked updates. The most recent was \"Harness guide: what happens after a coding agent opens a pull request\" on 2026-11-20.","window":{"days":30,"updateCount":54,"returned":50},"generatedAt":"2026-11-22T07:10:20.529Z","updates":[{"title":"Harness guide: what happens after a coding agent opens a pull request","summary":"Harness published a guide on November 20, 2026, by Mrinalini Sugosh, arguing that a coding agent opening a pull request is only the midpoint of an autonomous software delivery lifecycle, and that what follows determines whether AI-written code ships safely. The article defines an autonomous SDLC as a process in which AI agents write, test, secure, review and ship changes while policy gates enforce rules and humans approve risky steps. It describes a risk-based autonomy model with three tiers: autonomous for low-risk changes such as dependency bumps and dev-environment work; human-in-the-loop for standard production changes, new services and security fixes; and deterministic, fixed-step paths for the highest-risk areas such as payments, regulated data and destructive schema operations. Identity, policy, enforcement and evidence stay the same across all tiers. The piece ties this to Harness's Software Delivery Agent, which it says takes code, infrastructure and database changes to production under user-defined policies.\n\nThe article walks through seven stages (plan, generate, test, secure, review, deploy and verify), pairing each with what the agent does and what the gate checks. For example, the test gate requires that tests pass and that none were deleted, skipped or weakened, and the deploy gate uses policy as code covering approvals, freeze windows and required steps. After a pull request is opened, the article describes a sequence. The pipeline runs builds and tests, and Harness Builds uses Test Intelligence to run only the affected tests. Security scans for static analysis, dependency analysis and secrets detection run as pipeline stages. A Triage Agent deduplicates findings and prioritizes them using CVSS, EPSS and reachability analysis. A policy check using Open Policy Agent confirms that required checks and approvals exist. AI Code Review ranks the diff by risk and grounds its feedback in an SDLC Knowledge Graph that includes past incidents. Releases roll out via canary, blue/green or rolling strategies, and AI Verification compares live metrics and logs with the previous version, triggering rollback or alerts if performance is worse.\n\nThe guide also offers a red-flag checklist for human reviewers covering weakened tests, edited workflow and build files, duplicated utilities, new dependencies (including the risk of nonexistent package names that attackers could register) and exposed secrets. It proposes a rule that changes weakening checks, altering build or deployment, or adding dependencies, permissions or workflows require human approval, as do changes touching production paths, authentication, payments or regulated data. It cites GitHub guidance on the volume of agent pull requests and OWASP guidance that every AI-generated change needs a human owner who approved it. The article closes with a section asking whether AI agents can fix security vulnerabilities safely, answering yes under conditions that include scoped changes and tests as the gate. The source text is truncated before the final condition appears.","date":"2026-11-20","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/blog/what-happens-after-a-coding-agent-opens-a-pull-request-a-guide-to-the-autonomous-sdlc","publisher":"harness.io"},{"title":"Harness publishes layered GitOps security guide for AI-generated Kubernetes YAML","summary":"Harness published a blog post arguing that GitOps alone does not make AI-generated Kubernetes YAML safe and recommending a layered setup. The layers are CI validation via server-side dry runs, Harness OPA policies evaluating GitOps Applications, Kubernetes admission controls, and Harness AI Verify for post-deployment health checks. The post describes Harness GitOps as building on Argo CD with a centralized management layer for applications, repositories, and agents. It is dated November 20, 2026.","date":"2026-11-20","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/gitops-ai-generated-kubernetes-yaml-safety-layered-setup","publisher":"harness.io"},{"title":"Harness blog guide: zero-config Gradle remote caching with Build Intelligence","summary":"Harness published a guide explaining Gradle's build cache mechanics and how custom Java tasks can be made cacheable. It positions Harness CI's Build Intelligence as a zero-config, managed remote cache that the Harness agent configures automatically for Gradle, Maven 3.9+, Bazel, and Go. Enabling it requires a buildIntelligence flag at the pipeline stage level, and adding --profile to publish savings to the Harness UI.","date":"2026-11-19","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/blog/fast-ci-master-gradle-build-caching","publisher":"harness.io"},{"title":"Harness adds private connectivity and egress control to Cloud Runners","summary":"Harness has extended its managed Cloud Runners CI infrastructure with Private Connectivity, which lets builds reach private repositories, artifact registries, databases, and internal services without public endpoints, using a Tailscale-powered encrypted overlay and a customer-managed connectivity appliance. It also adds Outbound Egress Controls, powered by the Runtime Protection Agent, which let security teams allow or block outbound HTTP/HTTPS destinations by domain, URL, or path across Linux and Windows builds, with Private Connectivity also supporting macOS. The blog positions these as enabling enterprises to scale CI without running their own build fleets while meeting network security policies, and notes availability should be reviewed with a Harness account team.","date":"2026-11-19","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/enterprise-ready-harness-cloud-runners","publisher":"harness.io"},{"title":"Harness details nginx2root: five nginx RCE flaws and 2,372 exposed APIs","summary":"Harness published an analysis of nginx2root, five unauthenticated nginx RCE flaws disclosed November 18, 2026 by Xint researchers, led by CVE-2026-123007 in HTTP/2 code dating to 2015. Fixed releases are nginx 1.30.7 and 1.31.9, and Harness reports that its own discovery data shows 2,372 APIs across at least 11 customers reporting vulnerable nginx OSS versions, though HTTP/2 was not confirmed on those. The post also stresses that WAFs and DAST scans are largely unable to detect the exploit, and recommends restarting nginx after upgrading or disabling HTTP/2 and HTTP/3 as interim mitigations.","date":"2026-11-19","dateIsEstimated":false,"signalType":"security_incident","signalTypeLabel":"Security","sourceUrl":"https://www.harness.io/blog/nginx2root-five-nginx-rce-flaws-cve-2026-123007-and-how-to-patch-them","publisher":"harness.io"},{"title":"Harness publishes second edition of Next-Generation CI/CD For Dummies","summary":"Harness released a second edition of its CI/CD For Dummies eBook, dated November 18, 2026, covering how pipelines should handle AI-generated code, policy-as-code and RBAC governance, progressive deployment with automated verification and rollback, and ten capabilities to look for in a next-generation CI/CD platform. The guide is positioned as a starting point for developers, platform engineers, and engineering leads modernizing pipelines for AI-assisted development. It is delivered as a downloadable PDF after a form submission.","date":"2026-11-18","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/resources/next-generation-ci-cd-for-dummies","publisher":"harness.io"},{"title":"Harness blog positions Software Delivery Agent as Jenkins alternative for AI-driven PR volume","summary":"Harness published a blog post, dated November 18, 2026 and last updated November 17, 2026, by Lavakush Biyani, arguing that AI coding tools such as Cursor are increasing pull request volume and that self-hosted Jenkins pipelines sized for earlier workloads will queue jobs and slow developer feedback. The post describes Jenkins' controller-agent model as requiring teams to keep adding and maintaining build agents, plugins, Groovy Jenkinsfiles, and shared libraries, and presents the Harness Software Delivery Agent as an alternative for scaling CI. The piece is educational and product-positioning in nature rather than a new product announcement, and it describes capabilities presented as already available.\n\nThe capabilities listed are execution on Harness Cloud managed infrastructure, with self-managed and Kubernetes-based options for teams with specific infrastructure or compliance requirements. For build speed, the post cites Cache Intelligence, Build Intelligence, and Docker Layer Caching, which aim to avoid repeated dependency downloads, unnecessary recompilation, and full Docker image rebuilds. For testing, it describes Test Intelligence, which selects unit tests associated with a code change rather than running the full suite, along with parallel test execution. It also lists parallel stages and steps and matrix strategies, which let teams run the same workflow across multiple runtime versions or configurations without duplicating pipeline definitions. For governance, it names reusable pipeline templates, centralized secrets and connector management, role-based access controls, and governance policies intended to standardize pipelines across teams.\n\nThe post's most distinctive claim is about AI coding assistants. Through a Harness Model Context Protocol (MCP) server, assistants including Cursor, Windsurf, and Claude Code can work with CI/CD pipelines via natural-language instructions, provided the appropriate MCP integration and permissions are configured. Example prompts in the article include creating a pipeline that builds a project and runs its unit tests, triggering the CI pipeline for the current branch, and checking the latest pipeline execution.\n\nThe TL;DR condenses the message into four themes: scaling builds without managing every build machine, accelerating builds through caching, reducing CI feedback time through test selection and parallelism, and connecting AI coding tools to pipelines through MCP. The introduction frames the content as guidance for simplifying Jenkins migration, but the body focuses on feature descriptions and links to Harness developer documentation for each capability. The article does not include customer results, benchmark figures, pricing, or availability dates for any feature.\n\nCompetitively, the post ties Harness's CI offering to the rise of AI-assisted development and targets Jenkins users with a migration-oriented message. Its MCP integration with popular AI coding tools is presented as a differentiator. For competitors in CI/CD, it signals a content strategy focused on displacing self-hosted Jenkins and on owning the workflow between AI coding assistants and pipeline execution.","date":"2026-11-18","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/blog/is-your-jenkins-pipeline-ready-for-what-cursor-just-handed-you","publisher":"harness.io"},{"title":"Harness details CVE-2026-114938 JVM Base64 decoder out-of-bounds write","summary":"Harness researcher Roshan Piyush published lab results on CVE-2026-114938, a memory-corruption bug in the JVM's AVX-512 Base64 intrinsic. A malformed pre-auth value makes the fast decoder write 62 bytes past a heap array, returning a normal 401 response. Under default heap settings the write is silent, but with small fixed TLABs the lab reproduced JVM crashes (SIGSEGV in G1), so denial of service is demonstrated while code execution is not. The post cites Oracle's November 17, 2026 Critical Patch Update covering 11 Java SE issues and lists replacement builds, and recommends inventorying JVMs, upgrading, or temporarily disabling the intrinsic with -XX:-UseBASE64Intrinsics.","date":"2026-11-17","dateIsEstimated":false,"signalType":"security_incident","signalTypeLabel":"Security","sourceUrl":"https://www.harness.io/blog/past-the-end-of-the-buffer-cve-2026-114938","publisher":"harness.io"},{"title":"Harness hosts on-demand webinar on the Autonomous SDLC","summary":"Harness published an on-demand webinar, 'The Autonomous SDLC: Moving from Point Tools to a Unified DevSecOps Platform', arguing that fragmented point tools block AI agents and automation and that a unified platform provides the needed context. The session covers context gaps, AI-driven pipeline decisions, and the platform advantage over fragmented CI/CD toolchains. Page metadata lists a November 17, 2026 publication date.","date":"2026-11-17","dateIsEstimated":false,"signalType":"event","signalTypeLabel":"Event","sourceUrl":"https://www.harness.io/resources/the-autonomous-sdlc","publisher":"harness.io"},{"title":"Harness hosts on-demand webinar on knowledge graphs for AI agents","summary":"Harness published an on-demand webinar on why AI agents need a semantic layer over enterprise data, with a live demo of the Harness Knowledge Graph for the software delivery lifecycle (SDLC). The session covers how the graph was built, what data it holds and can ingest, and use cases including coding agents. The page lists a publication date of November 17, 2026, while structured metadata gives November 18, 2026.","date":"2026-11-17","dateIsEstimated":false,"signalType":"event","signalTypeLabel":"Event","sourceUrl":"https://www.harness.io/resources/giving-ai-agents-context-the-power-of-knowledge-graphs","publisher":"harness.io"},{"title":"Harness previews AI agent knowledge graph for SDLC at virtual event","summary":"Harness revealed it has built a knowledge graph tailored to the software delivery lifecycle (SDLC) to provide AI agents with structured context about entities and relationships in development workflows. The company will demonstrate this capability in a live session on November 17, 2026, positioning it as a semantic layer critical for agentic AI in DevOps.","date":"2026-11-17","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/events/giving-ai-agents-context-the-power-of-knowledge-graphs","publisher":"harness.io"},{"title":"Harness publishes 2026 flaky-test tooling comparison, highlighting CI-native test management","summary":"Harness published an independent comparison of seven flaky-test tools in 2026, emphasizing that reruns alone are insufficient and advocating for quarantine policies and test-health visibility. The post positions Harness CI’s same-commit flaky detection, quarantine, and test-health policies as CI-native alternatives to dedicated reliability layers.","date":"2026-11-17","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/blog/flaky-test-tooling-landscape-2026","publisher":"harness.io"},{"title":"Harness survey: 700 engineering leaders overstate AI agent controls","summary":"A Harness blog post by Eric Minick draws on its State of Agent DLC 2026 report, which surveyed 700 engineering leaders. Leaders report 74-77% confidence in their AI agent controls across evals, security, inventory, cost, and rollback, but the actual safeguards trail by 30-55 points: only 19% have a blocking eval gate on every release, 87% had a security incident, only 44% run shadow AI discovery tooling, 60% overran AI budgets, and only 33% have a kill switch. The post offers three diagnostic questions for teams and encourages readers to download the full report.","date":"2026-11-17","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.harness.io/blog/confidence-without-control-the-ai-agent-confidence-gap-in-2026","publisher":"harness.io"},{"title":"Harness blog: symlinks let AI coding assistants bypass path permission checks","summary":"Harness published the third post in a security research series on AI coding assistants, written by Soujanya Namburi and Roshan Piyush, examining how permission checks can disagree with the system actions they gate. The central point is that a permission layer typically evaluates the path string an agent requests, while the kernel follows symlinks to the actual file, so a symlink committed to a repository can make a read that passes a workspace check open a file outside the workspace, such as a user's AWS credentials. In the authors' test, a policy check on the string workspace/docs/api-reference.md returned ALLOWED while the kernel resolved the read to /Users/you/.aws/credentials. The post notes that Claude Code has fixed this gap three times (CVE-2025-59829, CVE-2026-25724, CVE-2026-39861), that Tego reported a similar CLAUDE.md import issue in July, and that Anthropic reportedly called it informative because workspace trust is the boundary. It also cites Cato Networks' report of symlink-validation flaws in Cursor's sandbox (CVE-2026-50548 and CVE-2026-50549), fixed in Cursor 3.0, and a Codex CLI case where an allowlisted git show command wrote a malicious diff-tool entry through an output option. The authors argue these bugs recur structurally because values are transformed from text to path to file to process, and that time-of-check to time-of-use gaps mean even realpath checks can be defeated by a later file swap. Recommended fixes include opening the file first and checking the file descriptor, using openat2 with RESOLVE_BENEATH on Linux, and tying approvals to the opened file. Interim steps include auditing cloned repos for symlinks with find . -type l -ls, keeping deny rules while noting they match strings, and running agents in containers that do not mount ~/.aws or ~/.ssh, using short-lived credentials. The post closes with a teaser for a follow-up on configuration switches that disable safety checks.","date":"2026-11-17","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.harness.io/blog/ai-coding-assistants-the-disguise-in-plain-sight","publisher":"harness.io"},{"title":"Harness ties Microsoft 365 outage lessons to its Resilience Testing product","summary":"Harness published a blog post on November 16, 2026, using the November 12, 2026 Microsoft 365 outage as a case study for its Resilience Testing product. Per the post, Microsoft began investigating connectivity problems affecting Exchange Online and dependent Microsoft 365 features that day. Microsoft's incident updates, as cited by Harness, describe servers failing automated health checks, lost capacity shifting workload onto remaining infrastructure, a connection-establishing process becoming overwhelmed, servers becoming unresponsive again when traffic returned, and a network configuration change being reverted. The post notes that Microsoft tied that configuration change to the incident as a correlation rather than a confirmed root cause in the updates reviewed. Harness then maps four scenarios to its product set. Resilience Risk Insights, which Harness describes as AI-powered and read-only, analyzes Kubernetes manifests, deployment configurations, pipeline history, and infrastructure metadata to flag issues such as missing CPU and memory limits, insufficient resilience test coverage, and untested dependencies, producing a ranked Resilience Risk Score. Chaos Testing is presented for controlled fault injection such as pod termination, node drain, CPU and memory stress, and network disruption, and Load Testing can be combined with those experiments. Application topology mapping and blast-radius analysis are described for tracing cascading failures across dependent services. Disaster Recovery Testing is positioned to automate regional failover experiments and measure results against defined RTO and RPO targets, with audit evidence generated. The post is a thought-leadership and product-positioning piece rather than a product launch; it does not announce new features, pricing, or availability dates. The article is truncated before its conclusion, so any further claims are not captured here. For engineering and SRE audiences, the takeaway is Harness's framing that resilience should be validated before incidents, not only after restarts, and that its tooling is positioned against that need.","date":"2026-11-16","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/blog/microsoft-365-outage-how-harness-resilience-testing-helps-identify-and-prevent-similar-infrastructure-failures","publisher":"harness.io"},{"title":"Harness blog: AI coding cost attribution from tokens to features","summary":"Harness published a recap of an InfoQ webinar, recorded October 29, 2026, in which Harness Field CTO Martin Reynolds and Trimble's Barani Mohandoss argued that provider usage data alone cannot show AI coding value, and that spend should be rolled up to features and paired with velocity, quality, and cost metrics. The speakers described Trimble's use of Harness Engineering Efficiency to correlate AI spend with engineering output, and recommended role-based budgets, counter-metrics, and plan-then-execute model routing. The post also notes that forecasting token cost per feature remains unsolved.","date":"2026-11-16","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.harness.io/blog/ai-coding-cost-attribution-tokens-to-features","publisher":"harness.io"},{"title":"Harness adds agent authorizers for Bedrock AgentCore deployments","summary":"Harness published a blog post describing a new agent authorizer feature in its CD platform for Amazon Bedrock AgentCore agents. Teams can set either an AWS_IAM authorizer for callers inside AWS using SigV4 credentials, or a CUSTOM_JWT authorizer for external callers using OIDC/OAuth tokens from providers such as Okta, Auth0, Entra ID and Cognito, with rules on audiences, clients, scopes and custom claims. An optional Restrict to Gateway setting forces all calls through the AgentCore Gateway. The post also covers canary traffic shifting with automatic rollback, plus RBAC, OPA policies, approvals and audit trails for agent stages. The feature is gated behind the flag CDS_AGENT_RUNTIME_DEPLOYMENT, which requires contacting Harness Support to enable.","date":"2026-11-16","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/control-access-to-ai-agents-deployed-on-aws-bedrock-agentcore","publisher":"harness.io"},{"title":"Harness demonstrates AI-powered Kubernetes deployment debugging in new blog post","summary":"Harness published a technical blog demonstrating how its AI DevOps Agent (Harness AI) diagnosed and resolved a Kubernetes deployment failure in a single session. The post highlights AI-assisted debugging for pipeline configuration issues, missing database dependencies, and schema errors, reducing troubleshooting time from hours to minutes.","date":"2026-11-15","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/blog/debug-kubernetes-deployment-failures-with-ai","publisher":"harness.io"},{"title":"Harness shares ARTEX threat hunt findings across customer API traffic","summary":"Harness published a threat hunt analyzing ARTEX framework traffic across seven customer environments, identifying over 1,300 API spans in 30 days. The research highlights how endpoint, identity, and response correlation reveals potential risks, with observed patterns including public page crawls, API bursts, and denied requests to sensitive endpoints.","date":"2026-11-13","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/blog/an-artex-threat-hunt-across-customer-api-traffic","publisher":"harness.io"},{"title":"Harness FME adds Metric Check step to gate rollouts on live metrics","summary":"Harness introduced a Metric Check step in its Feature Management & Experimentation (FME) platform that automatically evaluates feature flag metrics against user-defined thresholds during pipeline runs. The step fails the pipeline if conditions are met, enabling automated rollback or stoppage based on live error rates or latency metrics tied to specific treatments.","date":"2026-11-13","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/fme-metric-check-step-gate-rollouts","publisher":"harness.io"},{"title":"Harness launches Infrastructure Efficiency Loop with IaC remediation agents","summary":"Harness introduced the Infrastructure Efficiency Loop, automating cloud waste remediation by converting cost and drift findings into IaC pull requests via new agents. The system includes a remediation agent for code changes, a blast radius agent for risk scoring, and cost visibility per workspace, all within Infrastructure Deployments.","date":"2026-11-12","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/close-the-loop-on-cloud-waste","publisher":"harness.io"},{"title":"Harness: Code quality is the new bottleneck in AI-first engineering","summary":"Harness argues that AI coding agents are outpacing validation, making code coverage a critical but insufficient quality signal. The company positions itself as guiding engineering leaders on embedding coverage and risk-based rules into CI workflows to automate validation and feedback for AI agents.","date":"2026-11-12","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/resources/code-quality-is-the-new-bottleneck","publisher":"harness.io"},{"title":"Harness launches Commitment Orchestrator for AWS Savings Plans and Reserved Instances automation","summary":"Harness introduced Commitment Orchestrator, a tool within its Cloud & AI Cost Management suite, to automate AWS Savings Plans and Reserved Instances (RIs) management. The feature enables six-step automation with guardrails, tranche-based purchasing, and tracking to optimize cloud spend while preventing over-commitment.","date":"2026-11-12","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/automate-aws-savings-plans-reserved-instances","publisher":"harness.io"},{"title":"Harness positions AI security platform as code-to-runtime leader in 2026","summary":"Harness argues that true AI security platforms must cover code, pipeline, and runtime—not just one stage—highlighting gaps in competitors' offerings. The post claims only 6% of organizations have fully automated continuous delivery, where security findings stall, and positions Harness as the only platform enforcing policies across the entire software delivery lifecycle.","date":"2026-11-11","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/blog/ai-security-platform","publisher":"harness.io"},{"title":"Harness enhances AI agent security with NVIDIA OpenShell integration","summary":"Harness introduced new AI security features to complement NVIDIA’s OpenShell architecture, including agent discovery, AI red teaming, API security testing, and runtime threat detection. The integration builds a context graph of agent dependencies and applies enforcement at runtime to prevent prompt injection, tool misuse, and data exposure.","date":"2026-11-10","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/securing-agents-and-their-apis-with-harness-and-nvidias-openshell-architecture","publisher":"harness.io"},{"title":"AI is only as powerful as the context behind it","summary":"Harness introduced a Software Delivery Knowledge Graph that provides AI agents with a continuously updated, relationship-aware model of delivery environments, integrating Google Cloud's Developer Connect. The system translates raw tool data into structured context for AI-driven pipeline generation, root-cause analysis, and cost anomaly tracing, reducing investigation time from 20 minutes to seconds.","date":"2026-11-09","dateIsEstimated":true,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/products/platform/knowledge-graph","publisher":"harness.io"},{"title":"Harness: AI coding assistants’ approval systems create security risks","summary":"Harness published a blog post analyzing how AI coding assistants cache approvals by name rather than content, enabling silent privilege escalations when project content changes. The post demonstrates proof-of-concept attacks on tools like Claude Code and Cursor, showing how attackers can swap benign components for malicious ones without re-triggering consent dialogs.","date":"2026-11-06","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/blog/ai-coding-assistants-the-approval-that-never-ends","publisher":"harness.io"},{"title":"Harness: Why AI Cost Management Isn't Cloud FinOps 2.0","summary":"Harness argues AI spend is growing twice as fast as cloud spend, with 25% wasted and no clear ownership, making traditional FinOps inadequate. The post highlights Harness's AI cost management tool, which tracks adoption, impact, and cost in one dashboard, enabling plain-language queries, anomaly detection, and multi-vendor budget enforcement.","date":"2026-11-06","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/blog/same-curve-twice-the-speed-why-ai-cost-management-isnt-cloud-finops-2-0","publisher":"harness.io"},{"title":"Harness Q3 FY27: Pipeline execution control, OIDC identities, and expanded policy governance","summary":"Harness Pipeline in Q3 FY27 introduces project-level concurrency partitions, selective stage execution for DAG pipelines, pipeline identities for OIDC authentication, and a customizable run form. Policy as Code expands with the ability to run policy sets on private infrastructure and Blast Radius Analysis for pre-save change impact previewing.","date":"2026-11-05","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/q3-fy27-product-update-harness-pipeline","publisher":"harness.io"},{"title":"Harness launches Software Delivery Agent to automate the outer loop of software delivery","summary":"Harness introduced the Software Delivery Agent, an AI-powered platform that automates the post-PR stages of software delivery—review, testing, security, and deployment—closing the gap between coding and production. The agent uses a four-layer architecture (interface, context, reasoning, execution) with shared permissions and audit trails, and includes a Trust Gate to prevent AI agents from gaming their own checks.","date":"2026-11-05","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.harness.io/blog/beyond-the-coding-agent-how-the-software-delivery-agent-closes-the-gap-to-production","publisher":"harness.io"},{"title":"Harness GitOps Q3 FY27: post-production rollback, PR pipeline enhancements, and Argo CD 3.5","summary":"Harness introduced post-production rollback from the Services dashboard, richer PR pipeline controls (e.g., GitHub status checks, automatic branch cleanup, pattern-based file path matching), Argo CD import history, agent metrics, and Argo CD 3.5 support. These updates improve visibility, control, and automation in GitOps workflows for Harness users.","date":"2026-11-05","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/q3-fy27-product-update-harness-gitops","publisher":"harness.io"},{"title":"Harness rebuilds platform for agent-driven delivery with unified pipeline model and Delegate","summary":"Harness introduced a rebuilt platform experience designed for autonomous software delivery with agents, consolidating build, deploy, security, infrastructure, and agentic work into a single pipeline model with flatter YAML and smart defaults. The update includes a unified Go-based Delegate for orchestration, versioned templates for controlled step adoption, and agents as a primary interface with enterprise governance controls.","date":"2026-11-05","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/introducing-a-new-platform-experience","publisher":"harness.io"},{"title":"Harness Release Orchestration Q3 FY27: semantic versioning, Git-backed processes, OPA governance, and fine-grained permissions","summary":"Harness shipped 12 enhancements to Release Orchestration in Q3 FY27, centered on better artifact tracking with semantic versioning, looping over targets, Git-backed processes, and stronger governance via OPA policies and fine-grained permissions. Key additions include default semantic versioning for artifacts, looping strategies for multi-target deployments, and bidirectional Git sync for release definitions.","date":"2026-11-05","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/q3-fy27-product-update-harness-release-orchestration","publisher":"harness.io"},{"title":"Harness Q3 FY27: Helm v4, Kubernetes visibility, ASG/ECS enhancements, and AI agent deployments","summary":"Harness introduced Helm v4 support with a new Dry Run step, live Kubernetes cluster and pod visibility, and Kubernetes Gateway API integration for traffic routing. AWS deployments gained warm pools for ASG, phased rollback, and ECS Blue-Green improvements. AI agent deployments now support importing existing agents, gateway management, and S3 artifact sources.","date":"2026-11-05","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/q3-fy27-product-update-harness-deployments","publisher":"harness.io"},{"title":"Harness publishes guide on securing AI agents with OWASP Top 10","summary":"Harness released a comprehensive guide applying the OWASP Top 10 for Agentic Applications across the software development lifecycle, emphasizing both development-time scanning and runtime protection for AI agents. The post highlights risks like supply chain vulnerabilities, tool misuse, and memory poisoning, and positions Harness as a leader in AI agent security.","date":"2026-11-04","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.harness.io/blog/ai-agent-security-applying-the-owasp-top-10-across-the-software-development-lifecycle","publisher":"harness.io"},{"title":"Harness integrates Code Context Engine and Software Delivery Knowledge Graph to improve AI coding agents","summary":"Harness explains why AI coding agents lose context after opening a PR and introduces a solution combining the Code Context Engine (from Augment Code acquisition) with the Software Delivery Knowledge Graph. This integration provides agents with both code and delivery context to reduce risks in the SDLC.","date":"2026-11-04","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/why-ai-coding-agents-lose-context-after-pr","publisher":"harness.io"},{"title":"Harness introduces AI Budgets to cut AI cost per million tokens by 55%","summary":"Harness revealed AI Budgets, a governance feature that reduced its own AI spend by 55% per million tokens while maintaining developer velocity. The tool enforces live budgets, graduated alerts, and self-service approvals, shifting from policy documents to automated controls.","date":"2026-11-03","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/events/cutting-ai-cost-per-million-tokens","publisher":"harness.io"},{"title":"Harness launches AI Budgets, cuts own AI cost per million tokens by 55%","summary":"Harness revealed it transformed its AI cost policy into a live, enforced system called AI Budgets, reducing its own AI spend per developer from $184 to $115 weekly and cutting cost per million tokens by 55%. The approach uses weekly budgets, graduated alerts, and self-service approvals to balance governance with developer productivity.","date":"2026-11-03","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.harness.io/resources/harness-on-harness-october-2026","publisher":"harness.io"},{"title":"Harness: Survey reveals daily delivery failures in legacy CI/CD, pushing autonomy agenda","summary":"A Harness-hosted webinar and guide argue that high deployment frequency in legacy CI/CD systems masks instability, off-hours rework, and burnout, especially as AI accelerates development. The content positions Harness’s autonomous delivery vision as the antidote to manual bottlenecks.","date":"2026-11-03","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/resources/fast-isn-t-enough-when-data-shows-daily-delivery-is-broken","publisher":"harness.io"},{"title":"Harness launches Unified API Inventory for cross-source API correlation","summary":"Harness introduced Unified API Inventory, a feature within its API Posture capability, to correlate API discoveries across live traffic, code repositories, and documentation. This reduces the time to patch vulnerabilities by linking runtime findings to their source code and ownership, addressing disconnected inventory and slow remediation in API security workflows.","date":"2026-11-02","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/unified-api-inventory-reduces-the-time-to-patch-with-cross-source-correlation","publisher":"harness.io"},{"title":"Harness analyzes Binance breach as a trust failure in identity and signing security","summary":"Harness published a blog post framing the $6.36B Binance breach as a failure of identity and release-path trust, not cryptography, highlighting a compromised DevOps engineer's laptop as the likely entry point. The post emphasizes the need for stronger controls between developer endpoints and production signing systems.","date":"2026-11-02","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.harness.io/blog/the-binance-breach-was-a-trust-failure","publisher":"harness.io"},{"title":"Harness ships 150+ features in October 2026, led by Cosmos Software Factory Agent and FedRAMP 20x Class A","summary":"Harness delivered 150+ features in October 2026, including the new Cosmos Software Factory Agent that automates code creation from tickets to pull requests, MCP sign-in via identity providers, and FedRAMP 20x Class A compliance. Other updates include AI Verify monitoring integration, AI agent deployments to AWS AgentCore, and generally available Code Coverage enforcement.","date":"2026-11-02","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/shipped-in-october-2026","publisher":"harness.io"},{"title":"Harness explains web application security testing types, tools, and best practices","summary":"Harness published a detailed guide on web application security testing, emphasizing the importance of SAST, DAST, IAST, and SCA in catching vulnerabilities before production. The post highlights OWASP’s 2025 Top 10 findings, including broken access control and software supply chain failures, and advocates for pipeline-native security testing to reduce incidents.","date":"2026-10-30","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/blog/web-application-security-testing","publisher":"harness.io"},{"title":"Harness: From Tokens to Features — AI cost attribution framework for engineering teams","summary":"Harness published a webinar outlining an architecture to connect AI coding tool spend (tokens) to engineering outcomes like commits, pull requests, tickets, and delivered features. The session argues provider-level usage data lacks granularity and proposes a vendor-neutral measurement layer for evaluating AI tool ROI.","date":"2026-10-30","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.harness.io/resources/from-tokens-to-features","publisher":"harness.io"},{"title":"Harness details CI performance fixes: Test Intelligence, quarantine, and sharding","summary":"Harness published a lab study showing how its CI pipeline features—Test Intelligence, flaky-test quarantine, and parallel sharding—reduce pipeline time and improve reliability. The post highlights a 52x gap between raw test time and pipeline time, and demonstrates how quarantine policies and state isolation prevent flaky tests from blocking merges.","date":"2026-10-29","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.harness.io/blog/flaky-tests-slow-ci-pipeline","publisher":"harness.io"},{"title":"Harness achieves SOC 1, FedRAMP 20x Class A, PCI DSS, and HIPAA milestones","summary":"Harness announced four compliance milestones: its first SOC 1 audit, FedRAMP 20x Class A certification, and first PCI DSS and HIPAA Security Rule assessments. SOC 1 and FedRAMP apply to the Harness Autonomous SDLC Platform, while PCI DSS and HIPAA apply to the Runtime Protection Agent.","date":"2026-10-28","dateIsEstimated":false,"signalType":"legal_regulatory","signalTypeLabel":"Legal","sourceUrl":"https://www.harness.io/blog/announcing-soc-1-fedramp-pci-and-hipaa","publisher":"harness.io"},{"title":"Harness Chaos Carnival: AI-powered resilience testing event","summary":"Harness hosted Chaos Carnival, an event focused on AI-powered software resilience testing that bridges chaos experiments with autonomous incident handling. The event featured speakers from Harness, Okta, Adobe, Accenture, and other industry leaders, highlighting practical approaches to uncovering system vulnerabilities before outages occur.","date":"2026-10-27","dateIsEstimated":true,"signalType":"event","signalTypeLabel":"Event","sourceUrl":"https://www.harness.io/event/chaoscarnival","publisher":"harness.io"},{"title":"Harness hosts FinOps Excellence Summit with industry leaders","summary":"Harness organized a virtual FinOps Excellence Summit featuring on-demand sessions on cloud cost optimization, AI/ML spend management, and collaborative FinOps strategies. Industry speakers from United Airlines, Google Cloud, Forrester, and others shared tactics for reducing cloud costs and aligning engineering, finance, and operations teams.","date":"2026-10-27","dateIsEstimated":true,"signalType":"event","signalTypeLabel":"Event","sourceUrl":"https://www.harness.io/event/finops-excellence-summit","publisher":"harness.io"},{"title":"Harness hosts Developer Experience Summit 2026: Unlocking DevEx in the Age of AI","summary":"Harness hosted the Developer Experience Summit 2026, a virtual event exploring how AI transforms software delivery beyond code generation. The summit featured talks on measuring engineering productivity, scaling AI responsibly, and rethinking engineering models, with speakers from Google Cloud, Okta, Siemens, and Harness leadership.","date":"2026-10-27","dateIsEstimated":true,"signalType":"event","signalTypeLabel":"Event","sourceUrl":"https://www.harness.io/event/developer-experience-summit","publisher":"harness.io"},{"title":"Harness DevOps Modernization Summit: AI-driven DevOps insights and governance","summary":"Harness hosted the DevOps Modernization Summit, featuring talks on AI-driven DevOps, scaling practices, and governance. Sessions included keynotes on AI’s impact, breakout tracks on AI code governance, infrastructure speed, and agentic systems, with Harness executives and external leaders like Avery Pennarun (Tailscale) and Matthew Skelton (Conflux).","date":"2026-10-27","dateIsEstimated":true,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.harness.io/event/devops-modernization-summit","publisher":"harness.io"}],"attribution":{"source":"Spyingbee","url":"https://spyingbee.com/updates/harness","citation":"Spyingbee, \"Harness updates\", https://spyingbee.com/updates/harness (retrieved 2026-11-22)"}}