K8sGPT changelog: what's new each month
Every stable K8sGPT release summarised by month: the highlights, new features, improvements, fixes and anything you need to act on. 12 months covered; the current month updates daily.
October 2026
2 releases: 0.4.40 → 0.4.41
October brought improvements to analyzer coverage and reliability, with better handling of edge cases across Kubernetes resource types and AI backend integrations.
Highlights
- Added ResourceClaim analyzer to detect missing DeviceClasses.
- Support for Claude 4.5 and 5.x models on Amazon Bedrock.
- Custom headers can now be sourced from the config file.
- Auto-pull missing ollama models on demand.
- Improved resilience against empty AI responses and connection hangs.
New
- ResourceClaim analyzer for missing DeviceClasses
- Claude 4.5 and 5.x model support for Amazon Bedrock
- Custom headers configuration in config file
- Auto-pull prompt for missing ollama models
- S3 object listing pagination
Improved
- Custom analyzer gRPC connections now close after each analysis
- StatefulSet start ordinals are now honored
- K8SGPT_BACKEND environment variable is honored on every serve start
- Empty AI responses and configured language are now properly handled
- Deployment replica mismatches in progress are ignored
- KEDA scaleTargetRef unsupported cases no longer reported as missing
- NetworkPolicy pod selector matchExpressions are honored
- Namespace filter now applied in Gateway API analyzers
Fixed
- API doc lookups for webhook analyzers corrected
- Empty storageClassName in StatefulSet claim templates now accepted
- Completion calls hardened against empty responses and hangs
September 2026
2 releases: 0.4.38 → 0.4.39
September brought resource-level analysis improvements and enhanced security posture. The month focused on finer-grained diagnostics and safer handling of Kubernetes objects.
Highlights
- Analyze individual resources with the new --resource flag for targeted troubleshooting.
- ValidatingAdmissionPolicy analyzer added to cover Kubernetes policy validation.
- Resource names anonymised in event-derived failures for improved privacy.
- Security updates to gRPC dependencies.
- Improved event filtering to reduce false positives.
New
- --resource flag to analyze a single resource instead of entire namespaces
- ValidatingAdmissionPolicy analyzer for Kubernetes admission policy validation
- Resource name anonymisation in event-derived failures
Improved
- Service events now filtered by involved object kind to reduce noise
- Namespace validation performed before running core analyzers
- Duplicate reporting of unsupported scaleTargetRef and missing references eliminated
- OpenAIClient guarded against empty completion choices
Fixed
- Security vulnerability in google.golang.org/grpc updated to v1.83.2
- Flag description corrected for --with-stat
August 2026
1 release: 0.4.37
August brought new AI provider support and stability improvements. The release focused on expanding integration options and fixing edge cases in Kubernetes resource analysis.
Highlights
- Added Amazon Bedrock Mantle as a new AI provider with OpenAI-compatible API.
- Integrated LiteLLM backend for broader language model support.
- Fixed concurrency handling in custom analysis to prevent deadlocks and panics.
- Improved robustness of Ingress and HTTPRoute analyzers against nil values.
New
- Amazon Bedrock Mantle AI provider support
- LiteLLM backend integration
Improved
- Max-concurrency clamping in custom analysis to prevent deadlock conditions
- Ingress analyzer now handles resource backends safely
- HTTPRoute analyzer guards against nil backend port values
- Deployment analyzer handles nil spec.replicas gracefully
Fixed
- Panic when processing Ingress resources with resource-type backends
- Nil pointer dereference in HTTPRoute backend port checking
- Nil pointer dereference in deployment replica specification
- Potential deadlock and panic in custom analysis with high concurrency
July 2026
2 releases: 0.4.35 → 0.4.36
July brought stability improvements and fixes across pod scheduling detection, gateway analyzers, and AI model integrations. Several dependency updates and bug fixes ensure more reliable cluster diagnostics.
Highlights
- Added detection for SchedulingGated pods to improve visibility into pod scheduling issues.
- Fixed handling of empty conditions in Gateway and GatewayClass analyzers to prevent crashes.
- Corrected Vertex AI model IDs and fixed parameter handling for Anthropic models.
- Protected cache operations against nil gRPC connections for more robust operation.
- Resolved nil reference issues in Service endpoint analysis.
New
- Detection for SchedulingGated pods to identify pods stuck in scheduling gates.
Improved
- Gateway and GatewayClass analyzers now guard against empty conditions.
- Service not-ready endpoints analysis now handles nil TargetRef safely.
- Vertex AI integration corrected to use proper model identifiers without stray characters.
- Anthropic model parameter handling now sends temperature or topp exclusively as required.
- Cache Store and Load operations protected against nil gRPC connections.
Fixed
- SchedulingGated pod detection was missing from pod analysis.
- Crashes in Gateway/GatewayClass analyzers when conditions were empty.
- Nil reference errors in Service not-ready endpoints handling.
- Vertex AI model IDs contained stray asterisks causing authentication issues.
- gRPC cache operations could fail on nil connections.
June 2026
1 release: 0.4.34
June brought expanded AI provider support and improved Azure configuration flexibility, along with several security and dependency updates.
Highlights
- Added support for Anthropic API as an AI provider option.
- Azure OpenAI users can now override the API version to use different Azure deployments.
- Security updates applied to AWS SDK and KEDA dependencies.
New
- Anthropic API support for AI-powered diagnostics.
Improved
- Azure OpenAI API version can be overridden for greater deployment flexibility.
Fixed
- Security update to AWS SDK for Bedrock Runtime.
- Security update to KEDA dependency.
May 2026
1 release: 0.4.33
May brought security updates and improvements to container restart analysis and ingress handling. The project also prepared governance documentation for CNCF incubation.
Highlights
- Enhanced analysis of logs from restarted containers for better diagnostics
- Security update to Ollama module dependency
- Fixed handling of empty TLS secret names in ingress resources
Improved
- Container restart log analysis now examines previous logs for better context
Fixed
- Empty ingress TLS secret names no longer cause analysis issues
April 2026
1 release: 0.4.32
April brought expanded cloud provider support and new Kubernetes resource analyzers to K8sGPT. The release adds capabilities for Azure and Amazon Bedrock while improving detection of common configuration patterns.
Highlights
- Added Azure API type support with custom HTTP headers for Azure environments.
- New DaemonSet analyzer with enhanced Pod and Job analysis for edge cases.
- Improved ConfigMap detection for sidecar patterns and GKE built-in ingress classes.
- Amazon Bedrock Claude models now support temperature and top-p parameters.
- Security update to gRPC dependency.
New
- Azure API type support with custom HTTP header configuration.
- DaemonSet analyzer for analyzing DaemonSet-specific issues.
- Recognition of GKE built-in ingress classes 'gce' and 'gce-internal'.
- Temperature and top-p parameter support for Amazon Bedrock Claude models.
Improved
- ConfigMap usage detection now handles sidecar patterns more accurately.
- Pod analyzer with additional special cases for edge conditions.
- Job analyzer with enhanced handling of special scenarios.
Fixed
- Security vulnerability in gRPC dependency (updated to v1.79.3).
March 2026
1 release: 0.4.31
March brought minor maintenance updates to K8sGPT, focusing on documentation and build consistency.
Improved
- Go version alignment with go.mod toolchain for consistent builds
February 2026
3 releases: 0.4.28 → 0.4.30
February brought new LLM provider support and security updates to K8sGPT. The release cycle focused on reliability improvements and compatibility fixes across multiple components.
Highlights
- Groq added as a supported LLM provider option
- OpenAI integration updated to use current token limits API
- MCP server handlers improved with better error handling and pagination
- Security updates applied to Prometheus and dependencies
New
- Groq LLM provider support
- Short flag and environment variable for metrics port configuration
Improved
- MCP server handlers now include better error handling and pagination
- JSON marshaling for custom REST prompts improved to handle special characters
- OpenAI integration switched to MaxCompletionTokens from deprecated MaxTokens parameter
Fixed
- Namespace validation added before running custom analyzers
- Security vulnerabilities addressed in Prometheus dependency
December 2025
1 release: 0.4.27
December brought an update to the MCP protocol stack, advancing K8sGPT's integration capabilities.
Highlights
- Updated to MCP v2 for improved protocol compatibility and performance
Improved
- MCP v2 integration for enhanced model context protocol support
October 2025
1 release: 0.4.26
October brought a minor bug fix release to K8sGPT.
Fixed
- Fixed missing filter argument in serve command
September 2025
1 release: 0.4.25
September brought a focused release addressing a critical inference issue that was preventing the agent from functioning correctly.
Highlights
- Fixed broken inference that was blocking core functionality
Fixed
- Inference system restored to working order
Summaries are written automatically from the official release notes (full changelog ↗); check the original notes before relying on a detail. K8sGPT: pricing, features and alternatives · All changelogs