Lightrun Release Notes🔗
Here's where you'll find the latest Lightrun release notes, including the main highlights, enhancements, and fixes made to each version as it is released.
For a monthly, high-level overview of major product updates, see the What's New page — a curated summary of the most significant features released each month.
Lightrun release cadence
Lightrun follows a consistent and predictable release cadence, enabling organizations to choose between early access to new features and more thoroughly tested versions.
The release flow differs based on the deployment type:
- Cloud-first releases: A new release is deployed to the Lightrun Cloud every two weeks. This environment receives the latest features, enhancements, and bug fixes first.
- Single-tenant and On-premise releases: These environments receive the same updates one week after the Cloud release, following additional validation and packaging.
- Quarterly Support Releases (QSRs): Once per quarter, a version is designated as a Quarterly Support Release. These versions are explicitly marked as QSR in the release notes (e.g., Version 1.55 QSR) and are maintained with critical security patches and fixes for high-impact issues until the next QSR becomes available. These version undergo extended backward compatibility testing and are ideal for environments that require longer upgrade cycles.
Backward compatibility
The Lightrun server supports agents and plugins released up to 12 months prior the server release. This only includes agents and plugins with the same major version as the server.
Additional resources
-
For release notes of previous releases, see:
- Functionality changes and deprecations: A consolidated list of feature changes, parameter changes, and deprecations across releases. Each change is also documented individually in the release notes. Before upgrading to a new version of Lightrun, we recommend reviewing this list of changes as it may include breaking changes that require your attention.
- Known Issues: For a list of known issues.
Version 1.89 [QSR]🔗
Release date: 15 July, 2026 (SaaS, Single-Tenant, and On-Premises)
Feature Enhancements🔗
MCP: Asynchronous Custom Metric Tools
Lightrun MCP adds an asynchronous tool set for custom metrics: custom_metric_create, custom_metric_status, custom_metric_cancel, and custom_metric_samples.
This model separates action creation from status polling and sample retrieval, so AI assistants can collect numeric runtime metrics without blocking on a single tool call.
For more information, see Supported tools.
MCP: Asynchronous Execution Count Tools
Lightrun MCP adds an asynchronous tool set for execution counters: execution_count_create, execution_count_status, execution_count_cancel, and execution_count_samples.
AI assistants can measure how often a line of code runs across agents over a sampling window, then retrieve status and samples asynchronously.
For more information, see Supported tools.
MCP: Threshold-Triggered Snapshots for Execution Duration
Lightrun MCP can capture a snapshot when measured execution duration exceeds a threshold you configure, including expressions, a maximum hit count, and the captured execution duration.
This capability applies to Java/JVM applications and requires a compatible Lightrun Java agent. It does not apply to Counter or Custom Metric actions.
For more information, see Supported tools and Metrics.
Service Configuration: Hide Default Agent Pool in IDEs
Company Admins can hide the Default Agent Pool from JetBrains IDEs and Visual Studio Code with the showDefaultAgentPoolInIde service setting. When the setting changes, IDEs refresh their agent-pool selection so developers work against the pools you intend them to use.
For more information, see Service Configuration.
Agent Pools: Threshold Exceeded Audit Event
Lightrun records an audit event when an agent pool exceeds its configured active agent threshold.
Java Agent: Java 26 Class-File Compatibility
The Java agent adds support for Java 26 class-file version 70 through ASM 9.10.1, improving compatibility with newer bytecode formats.
Note
This release improves class-file compatibility with Java 26. It does not by itself establish complete Java 26 runtime certification.
Security: Certificate Pinning Without Private Key
The certificate-pinning workflow no longer uses a TLS private key. Self-hosted deployments that use certificate pinning must provide valid SHA-256 hashes through the pinned-hash configuration. Lightrun validates configured certificate hashes at startup.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server, plugins, and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server, Keycloak, and related components.
Version 1.88 🔗
Release date: 30 June, 2026 (SaaS, Single-Tenant, and On-Premises)
Feature Enhancements🔗
Public API: Snapshot Hit Results
Lightrun exposes public API endpoints for retrieving snapshot hits for an action and snapshot hit data by ID, so you can integrate snapshot output into automation and external tooling. Requires DEV-level API permissions.
For more information, see Get snapshot action hits and Get snapshot hit data.
Agent Pools: Active Agent Threshold
You can set an active agent threshold on an agent pool in the Management Portal. When the number of active agents in the pool exceeds that threshold, Lightrun surfaces a warning so you can spot unexpected agent growth or licensing pressure in the pool. The threshold does not block agents from connecting.
For more information, see Manage Agent Pools.
MCP: get_actions Tool
Lightrun MCP adds the get_actions tool so AI assistants can find MCP-created actions from an ongoing or earlier investigation. Filter by agent pool, file, line, status, time range, or correlationKey, then use the returned actionId with the matching status and result tools (for example, snapshot_status or execution_duration_samples).
For more information, see Supported tools.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server, plugins, and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server, Keycloak, and related components.
Version 1.87 🔗
Release date: 21 June, 2026 (SaaS, Single-Tenant, and On-Premises)
Feature Enhancements🔗
Metrics: Threshold-Triggered Snapshots
For Method Duration and Tic & Toc metrics, you can configure Lightrun to automatically capture a snapshot when measured duration exceeds a threshold you define. This combines timing data with snapshot context—local variables and source location—so you can investigate slow executions without adding a separate snapshot action.
In the JetBrains plugin, select Capture a snapshot when threshold is exceeded, then set the threshold in milliseconds and the maximum number of snapshot hits. Threshold breaches are marked on the metric chart in the results console.
This capability is available from Lightrun version 1.87 in the JetBrains plugin for Java/JVM applications and requires Lightrun agent version 1.87 or later. It does not apply to Counter or Custom Metric actions.
For more information, see Metrics.
MCP: Asynchronous Execution Duration (Tic & Toc) Tools
Lightrun MCP adds an asynchronous tool set for measuring execution duration over a live code section (Tic & Toc): execution_duration_create, execution_duration_status, execution_duration_cancel, and execution_duration_samples.
This model separates action creation from status polling and sample retrieval, so AI assistants can wait for runtime executions without blocking on a single tool call. When the async tool set is available, the synchronous get_runtime_execution_duration and get_runtime_execution_duration_samples tools are hidden.
For more information, see Supported tools.
Public API: Log Messages Retrieval
Lightrun exposes a public API endpoint for retrieving log messages in a structured, machine-readable format, so you can integrate log output into automation and external tooling. Requires DEV-level API permissions.
For more information, see Get log messages.
Java Agent: Tic & Toc End Line Included in Duration
Tic & Toc measurements now include the end line (Toc) in the measured duration, giving more accurate block-duration readings when the last line performs meaningful work.
MCP: Ignore-Quota Support for snapshot_create
OAuth-authenticated users with ignore-quota permission can set ignoreQuota on snapshot_create to bypass quota limits when needed. API-key clients and users without permission receive a clear error.
For more information, see Supported tools.
MCP: Connect from IDE (JetBrains)
The JetBrains plugin can configure AI assistants installed on your machine to connect to Lightrun MCP from the IDE, so you can set up OAuth-based MCP access without editing config files manually. This applies to any MCP-compatible agent you run locally—for example Claude Code, Cursor, GitHub Copilot, and Amazon Q.
For more information, see MCP quickstart.
Lightrun AI: Ask Prod Skill
Lightrun adds the Ask Prod Skill — a reusable workflow that guides AI assistants to answer production questions using live runtime evidence from Lightrun MCP (values, durations, counts, and branch behavior).
For more information, see Ask Prod Skill.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server, plugins, and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server and Keycloak.
- Fixed an issue where Tic & Toc metrics created in VS Code could not be edited in the JetBrains plugin.
Version 1.87.1🔗
Release date: 28 June, 2026 (SaaS and Single-Tenant)
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server, plugins, and agents.
Version 1.86 🔗
Release date: 3 June, 2026 (SaaS, Single-Tenant, and On-Premises)
Feature Enhancements🔗
Management Portal: New Home Page
The Lightrun Management Portal has a new default home page that helps you get started with Lightrun AI, connect MCP, install Lightrun AI Skills, and reach setup and learning resources from one place. The existing Learn & Explore page remains available and is linked from the portal.
Java Agent: Wildcard Support in Extra Class Path
Starting with this release, lightrun_extra_class_path supports * and ? wildcards in paths and filenames—for example, /path/version*/arch*/*.jar—so versioned JARs are included without updating the path on every build. Wildcard entries must resolve to files, not directories.
For configuration details, see Customize the Lightrun Java agent.
MCP: Organization-Level MCP Access Control
Company Admins can control organization-wide Lightrun MCP availability with the MCP access toggle in Service Configuration. When disabled, Lightrun blocks MCP for all OAuth users and API-key-based clients in the organization.
For more information, see Lightrun MCP overview.
MCP: Split Runtime Source Discovery into Two Tools
Lightrun MCP replaces the get_runtime_sources tool with get_runtime_sources_agent_pools and get_runtime_sources_agents to make the source discovery phase more token-efficient. AI assistants retrieve agents only for the desired agent pool, with multiple filters and pagination, instead of receiving one large combined response.
For more information, see Supported tools.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server, plugins, and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server and Keycloak.
Version 1.85 🔗
Release date: 19 May, 2026 (SaaS, Single-Tenant, and On-Premises)
Feature Enhancements🔗
User Management: Group Membership Summary
The User Management table now includes a group membership summary, making it easier for administrators to quickly review user access context.
Java Agent: Dedicated Directories for Temporary Files and Logs
Starting with this release, you can configure Lightrun-specific directories for Java agent classpath indexer temporary files and agent logs using classpath_indexer_temp_dir and agent_log_target_dir. This lets you avoid relying on the operating system /tmp folder—useful in containerized environments and hosts where /tmp has limited space or strict cleanup policies—without changing the JVM-wide java.io.tmpdir setting.
For configuration examples and property details, see Reducing Java agent temporary file usage in /tmp.
MCP: Asynchronous Snapshot Tools
Starting with version 1.85, Lightrun MCP supports asynchronous snapshot tooling with a multi-step action lifecycle: snapshot_create, snapshot_status, snapshot_cancel, snapshot_get_values, and snapshot_get_call_stack.
This model separates action creation from status polling and result retrieval, so AI assistants can handle delayed or intermittent runtime hits more reliably. For the snapshot capability, the previous synchronous flow is hidden when the async tool set is available.
For more details, see Supported tools.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server, plugins, and agents.
Version 1.84 [QSR]🔗
Release date: 5 May, 2026 (Single-Tenant, and On-Premises to follow next week)
Feature Enhancements🔗
MCP: Improved Tool Call Success Rate
Resolved an MCP reliability issue where tool calls could time out without returning hits, even when hits were captured successfully. This fix significantly improves tool call success rates.
VS Code Extension: Capture More
Capture more is now supported in Visual Studio Code for Lightrun Snapshots. When snapshot data is limited by size, depth, or number of elements, you can use Capture more to collect full or deeper variable values on the next snapshot hit.
For more information, see Capture full and deeper snapshot values.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server.
Version 1.84.3🔗
Release date: [TBD] (On-Premise)
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server, plugins, and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server.
Version 1.84.4🔗
Release date: [TBD] (On-Premise)
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server.
Version 1.83 🔗
Release date: 26 April, 2026 (Single-Tenant, and On-Premises to follow next week)
Feature Enhancements🔗
JetBrains Plugin: Capture More for the Node.js Agent
Capture more is now supported for the Node.js agent in JetBrains IDEs. When snapshot data is limited by size, depth, or number of elements, you can click Capture more to create a new snapshot that captures the full or deeper value on the next snapshot hit.
For more information, see Capture full and deeper snapshot values.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server, plugins, and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server.
Version 1.82 🔗
Release date: 7 April, 2026 (Single-Tenant, and On-Premises to follow next week)
Highlights🔗
MCP: Temporary-Access Agent Pools in Get Runtime Sources
The get_runtime_sources MCP tool now includes agent pools to which the user was granted temporary access, not only agent pools available through persistent access.
JetBrains Plugin: Results Badge
The JetBrains plugin adds a results badge that makes it easier to see when new Lightrun results are available from your actions.
MySQL: Supported version 8.4.3
MySQL 8.0.38 has reached end-of-life and is no longer supported for Lightrun deployments. Starting with Lightrun Server 1.82.x and Helm chart 3.41.x, Lightrun supports MySQL 8.4.3.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server, agents, and plugins.
- Resolved security vulnerabilities (CVEs) in the Lightrun server, agents, and plugins.
Version 1.81 🔗
Release date: 28 March, 2026 (Single-Tenant, and On-Premises to follow next week)
Highlights🔗
Custom Metrics: Double Expressions
The custom metric action now supports metric expressions that evaluate to double, in addition to int.
For more information, see Custom metrics.
Java Agent: Java 26 Forward Compatibility
Java 26 is certified for forward compatibility: code written for earlier Java language versions runs correctly with Lightrun when executed on a Java 26 JVM.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server, plugins, and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server.
Version 1.80 🔗
Release date: 9 March, 2026 (SaaS, Single-Tenant, and On-Premises to follow next week)
Highlights🔗
MCP: New Tool — Count Runtime Executions
Lightrun MCP now includes a new tool that counts how many times a specified line of code runs across agents over a sampling window, so AI assistants can reason about real traffic and flow without code changes or redeployments.
For more details, see Supported tools.
MCP: API Key Authentication
You can authenticate Lightrun MCP using an API key, which enables organizational AI agents and automated systems to connect securely without interactive login flows.
For background and setup, see OAuth vs API key in the Lightrun MCP overview and AI agents and systems (API key) in the MCP quickstart.
JetBrains Plugin: Results Console Refresh
The Lightrun Results area in the JetBrains plugin has been refreshed to make it easier to find action outputs after creating a Snapshot, Trace, Metric, or Log. Result consoles now appear as tabs in the tab bar, with an updated filter bar design.
Metrics are now unified into a single Metrics tab with improved search, a Metric Type filter, resizable columns, and clearer Custom Metric expression display.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server.
Version 1.80.1🔗
Release date: 15 March, 2026 (SaaS, Single-Tenant, and On-Premises to follow next week)
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server and agents.
Version 1.79 🔗
Release date: 23 February, 2026 (Single-Tenant and On-Premises to follow next week)
Highlights🔗
MCP: New Tools — Get Runtime Callstack and Get Runtime Numeric Metric
Lightrun MCP now includes two new tools for AI assistants to inspect live runtimes:
- Get Runtime Callstack: Enables the LLM to retrieve the call stack at a specified code location, helping to understand execution flow and diagnose issues in context.
- Get Runtime Numeric Metric (+ samples): Enables the LLM to collect numeric metric samples from a specified code section across agents. The metric can be based on any code expression that evaluates to an integer number. Using the runtime dataset, the LLM can analyze trends, validate behavior, and generate reports.
For more details, see Supported tools.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server.
Version 1.78 [QSR]🔗
Release date: 17 February, 2026 (Single-Tenant, and On-Premises to follow next week)
Feature Enhancements🔗
JetBrains Plugin: Get Full and Deeper Snapshot Values
Starting with this release, we've added the Capture more feature to our Java agent. This feature will be rolled out to all other agents in upcoming releases.
When snapshot data is limited by size, depth, or number of elements, you can now click Capture more to create a new snapshot that captures the full or deeper value on the next snapshot hit. This removes the need to manually create watch expressions while preserving previously captured data.
For more information, see Capture full and deeper snapshot values.
MCP: New Tool — Get Runtime Execution Duration
The new Execution Duration tool enables the LLM to automatically collect execution-time samples for a specified code section across multiple agents. Using the full runtime dataset, the LLM can validate runtime behavior, identify anomalous patterns, and generate performance reports and dashboards. These tools are supported for Java only.
For more details, see Supported tools.
MCP: Certified AI Clients
Lightrun's MCP is designed to work out of the box with any AI coding assistant that supports the Anthropic MCP standard. In practice, some tools introduce minor deviations or non-standard connection flows. In this release, we continued an ongoing effort to validate and certify assistants one by one to ensure reliable, smooth connectivity to Lightrun's MCP.
Currently certified assistants: GitHub Copilot, Cursor, Claude Code, Codex, Amazon Q, Gemini Code, JetBrains AI Assistant, JetBrains Junie, Antigrabity, and Kiro.
For more information, see MCP Quickstart.
Plugin and Portal UX: Filter Actions by Origin
You can now filter Lightrun actions by their origin (MCP or Plugin) using the new Created In filter in the Lightrun Plugin or Management Portal. This makes it easier to focus on the actions that matter to you and reduces noise when working across multiple entry points.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server.
Version 1.78.2🔗
Release date: 11 March, 2026 (On-Premise)
Bug Fixes🔗
- Fixed internal bugs and security vulnerabilities in the Lightrun server.
Version 1.78.3🔗
Release date: 30 March, 2026 (SaaS, Single-Tenant, and On-Premises)
Highlights🔗
MySQL: Supported version 8.4.3
MySQL 8.0.38 has reached end-of-life and is no longer supported for Lightrun deployments. Starting with Lightrun Server 1.82.x and Helm chart 3.41.x, Lightrun supports MySQL 8.4.3.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server, plugins, and agents.
- Resolved security vulnerabilities (CVEs) in the Lightrun server.
Version 1.78.4🔗
Release date: 16 April, 2026 (On-Premise)
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server, plugins, and agents.
- Mitigated security vulnerabilities (CVEs) in the Lightrun server.
Version 1.78.5🔗
Release date: 5 May, 2026 (On-Premise)
Highlights🔗
MCP: Improved Tool Call Success Rate
Resolved an MCP reliability issue where tool calls could time out without returning hits, even when hits were captured successfully. This fix significantly improves tool call success rates.
Spring Boot: Upgraded to 3.5.14
The Lightrun server was upgraded to Spring Boot 3.5.14.
Bug Fixes🔗
- Resolved security vulnerabilities (CVEs) in the Lightrun server.
Version 1.77 🔗
Release date: 27 January, 2026 (SaaS, Single-Tenant, and On-Premises to follow next week)
Feature Enhancements🔗
MCP: Improved Tool Experience
Improved the tool descriptions to make the LLM understand how to more accurately use Lightrun’s tool and provide an overall better experience.
MCP: The LLM Can Now Configure How Many Expression Values to Get
We exposed a new parameter of the get_runtime_expression_value tool that allows the LLM to configure the maximum number of values to get within the range of 1-50 hits. The default value is 1. This parameter was previously hardcoded to 3. For more information, see MCP Overview Rules and Limits.
Traces: Enhanced Trace Creation Experience
Following your feedback, we’ve streamlined the Trace creation flow to reduce the number of steps required. You can now define and create Traces through a clearer, more efficient setup process. For more information, see Create a Dynamic Trace.
Bug Fixes🔗
- Fixed internal bugs in the Lightrun server.
- Fixed security vulnerabilities in the Lightrun server.
Version 1.76 🔗
Release date: 13 January, 2026 (SaaS, Single-Tenant, and On-Premises to follow next week)
Feature Enhancements🔗
All plugins: Plugin upgrade without server (On-Premises only)
Starting with version 1.76, admins can respond immediately to IDE changes, allowing developers to continue working with minimal disruption. Previously, the Lightrun server had to be updated by an admin for developers to get an updated plugin. Because server upgrades often occur infrequently, this resulted in delays, operational friction, and unnecessary downtime. As an admin, you upload an IDE plugin version through the Lightrun Management Portal. The uploaded file replaces the existing latest plugin file can be download from the Getting Started > **Install the plugin in your IDE* page, or obtained by contacting Lightrun Support. For more information, see Upload and distribute a Lightrun plugin independent of the server version.
Node.js Agent: Deploy Lightrun on Google Cloud Functions
Lightrun now supports deployment on Google Cloud Functions (Gen2) for Node.js, enabling real-time debugging and observability in serverless environments running on Cloud. This release includes a step-by-step deployment flow using the Lightrun Node.js agent, with support for deployment via the gcloud CLI or Cloud Build. For more information, see Deploy Lightrun on Node.js Google Cloud Functions.
Lightrun Management Portal: Find Your Lightrun Admin in Seconds
You can now view your organization’s Lightrun admins directly from the User Management page. View permissions have been extended to allow users to identify admins without requiring admin access.
Version 1.75 🔗
Release date: 4 January, 2026 (SaaS, Single-Tenant, and On-Premises to follow next week)
Highlights🔗
Lightrun MCP: Runtime Context for AI Coding Assistants
Lightrun MCP enables AI coding assistants such as Cursor, Gemini, and Copilot to inspect live application runtimes using Lightrun’s production-safe, on-demand capabilities.
Through the Model Context Protocol (MCP), AI agents can retrieve runtime information as needed—without requiring code changes, rebuilds, or redeployments. This allows AI tools to inspect call stacks, variable values, and execution flow precisely when and where issues occur.
Lightrun MCP operates via a MCP server embedded in the Lightrun server, that securely connects AI tools with running applications over HTTPS. The result is faster root-cause analysis of complex, non-reproducible issues directly within real runtime environments.
To get started, see: