8.9 Release notes
These release notes identify the main new features included in the 8.9 minor release, including alpha feature releases.
| Minor release date | Scheduled end of maintenance | Changelog(s) | Upgrade guides |
|---|---|---|---|
| 14 April 2026 | 13 October 2028 | Patch Releases and Changelogs | - |
- See release announcements to learn more about supported environment changes and breaking changes or deprecations.
- Refer to the quality board for an overview of known bugs by component and severity.
Technical Changelogs for all 8.9.x releases
Overview of all patch releases and their Changelogs in GitHub
8.9.0-alpha4
| Release date | Changelog(s) |
|---|---|
| 10 February 2026 |
Agentic orchestration
MCP Client connector operations and filter options
New operations are added in the Standalone mode to the MCP (Model Context Protocol) Client connector:
- List resources
- List resource templates
- Read resource
- List prompts
- Get prompt
In addition, the Tools panel in the element template is replaced by dedicated filter options for each operation.
- The Tools section in version 0 and 1 of the template is replaced by a Filters section in version 2.
- As Tool filters are not migrated during a Template Version upgrade, you must manually upgrade the template with any previously configured Tool filters.
Camunda 8 Run
Expanded RDBMS support
You can now configure Camunda 8 Run to use any of the supported secondary storage relational databases instead of the default H2.
This allows you to set up your local environment to match your production deployments, enabling deeper testing, faster debugging, and easier team onboarding.
Streamline your Camunda 8 Run experience
Camunda 8 Run is now easier to use with improved setup and configuration.
The CLI includes a helpful usage page, clearer error messages, especially for Elasticsearch startup, and prominently displays connection properties and credential information.
A revamped Java detection guided setup, log cleanup options, and better defaults for development environments (such as disk watermark thresholds) have been added. You can also start fresh using a new clean-state command, and the unified configuration file is now included and thoroughly documented.
Connectors
Fetch active process definitions
The connectors runtime now activates inbound connectors for earlier process versions if they have active instances waiting on message subscriptions. Previously, only inbound connectors from the latest process version were activated, which could prevent active instances of older versions from completing.
Console
New AWS US East region
With the new Camunda 8 SaaS AWS US East (us-east-2) region in North America, you can deploy orchestration workloads with full US data residency and improved regional stability.
Documentation
Camunda Docs MCP server
Use the Camunda Docs MCP server to query the latest official Camunda 8 documentation in your IDE or AI tool.
- If you use an AI coding tool such as Cursor or Copilot, add the MCP server to help ensure more accurate AI responses and code generation using Camunda 8 documentation and context.
- Once connected to the MCP server within your editor, you can ask context-aware questions about Camunda.
- The MCP server is available at the following URL:
https://camunda-docs.mcp.kapa.ai.
Modeler
Collaborate in Web Modeler
Live collaboration in Web Modeler is now more reliable with an improved collaboration experience.
Once you start editing a diagram, the canvas locks so only you can continue making edits to the diagram.
- Other users can view and interact with the diagram, but cannot make changes while it is locked by the current editor.
- Users with edit permissions can take over editing of the diagram by clicking Take over in the canvas lock bar.
This improvement creates a restricted but controlled collaborative environment, and helps prevent conflicts and broken sessions caused by multiple users editing the same diagram.
Import large process applications into Web Modeler
You can now import large process applications (containing a maximum of 100 supported files such as BPMN, DMN, forms, connector templates, and documentation) in a single step in Web Modeler or via direct import links in Self-Managed environments.
Import resources into Web Modeler
Shared global clipboard and duplicate elements
In Desktop Modeler and Web Modeler, you can copy-paste BPMN elements seamlessly between Camunda 7/8 diagrams and across clipboard-enabled BPMN tools/websites, and also quickly duplicate via Cmd+D (Mac) / Ctrl+D (Windows).
Tabs autosave in Desktop Modeler
The tabs autosave feature automatically saves diagram changes when you switch tabs in Desktop Modeler or when the app loses focus (for example, app switch, window blur). This ensures work is saved immediately, making it visible to your IDE and other tools without manual intervention.
Migration from Camunda 7 to Camunda 8
Audit log coverage for Camunda 7 to Camunda 8 migrations
The audit log migrator automatically converts Camunda 7 UserOperationLog entries to Camunda 8 AuditLog format, preserving the complete history of user operations including who performed actions, what entities were affected (process instances, tasks, variables, decisions), operation types (create, update, delete, assign, complete), timestamps, and annotations.
This ensures uninterrupted audit trail continuity across the migration, enabling customers to meet compliance requirements and maintain operational visibility without manual data reconstruction or workarounds.
Data Migrator history migration coverage
Orchestration Cluster
Schedule backups with the Orchestration Cluster
You can now configure scheduled backup intervals and retention directly in the Orchestration Cluster.
- External cron jobs are no longer needed.
- Supports setting duration schedules, manual ad‑hoc backups, API‑based updates, metrics, and audit logs.
- Backwards compatible with existing backup commands.
Unified configuration for the Orchestration Cluster
In Camunda 8.9, the remaining unified configuration project property changes are complete.
- All 8.9 property changes are documented in the Camunda 8.9 property changes table.
- You can search, sort, and filter the table to show breaking changes, direct mappings, and new properties.
- For more information on each property (including default values), see the property reference.
Property changes in Camunda 8.9
Only the first partial set of the unified configuration project properties was introduced in Camunda 8.8.
Process instance migration
Migrate from job-based user tasks to Camunda User Tasks
As part of process instance migration, you can now migrate active instances from legacy job‑based user tasks to modern, engine‑managed Camunda User Tasks via both the API and the Operate UI.
This lets you standardize on the Orchestration Cluster APIs and the recommended user task type before the removal of job‑based user task querying and management in the consolidated API.
RDBMS secondary storage
Continuous backup and restore
You can now back up and restore Camunda 8 when using RDBMS as secondary storage.
- Independent backup control plans handle primary and secondary backups separately while ensuring they align when restored.
- Disaster recovery is improved as you can recover Camunda instances with greater precision in data consistency.
Additionally, the new restore API syntax now supports --from and --to timestamp flags, enabling automatic selection of a compatible backup range.
- When no specific backup or timerange is specified, a restore is performed to the latest known position with no user interaction.
- Ensures version compatibility across backups and offers an override via
--allow-version-mismatch. - Reduces manual restore effort and enhances confidence in backup integrity, with reduced Recovery Time Objective (RTO).
8.9.0-alpha3
| Release date | Changelog(s) |
|---|---|
| 13 January 2026 |
Agentic orchestration
AWS Bedrock API key authentication support
The Amazon Bedrock model configuration now allows authentication using (long-term) Bedrock API keys as an alternative to the already existing authentication methods.
Model timeout configuration
The AI Agent connectors now support setting a timeout value on supported models.
Query parameters support on OpenAI compatible models
The OpenAI compatible model configuration now allows configuration of query parameters to be added to the model endpoint URL. This might be needed for custom API endpoints requiring additional metadata (such as API versions) to be set via query parameters.
Camunda 8 Run
Streamline your Camunda 8 Run experience
Camunda 8 Run is now easier to use with improved setup and configuration.
The CLI includes a helpful usage page, clearer error messages, especially for Elasticsearch startup, and prominently displays connection properties and credential information.
A revamped Java detection guided setup, log cleanup options, and better defaults for development environments (such as disk watermark thresholds) have been added. You can also start fresh using a new clean-state command, and the unified configuration file is now included and thoroughly documented.
Use H2 for data storage
Camunda 8 Run now includes H2 as the default secondary data store, providing:
- A lighter, simpler local development experience.
- Lower memory usage.
- A fully functional stack that doesn't require an external database.
New documentation shows you how to:
- Install Camunda 8 Run with H2 as the default secondary storage.
- Seamlessly switch from H2 to Elasticsearch or OpenSearch when required.
Cluster Metrics endpoint
Camunda 8.9 provides a Cluster Metrics endpoint for SaaS clusters.
- Activate a secure metrics endpoint for your cluster and integrate it with Prometheus, Datadog, or any monitoring system that supports Prometheus scraping.
- Get real-time visibility into cluster performance, troubleshoot faster, and integrate with your existing observability stack.
Global user task listeners
Camunda 8.9 introduces configuration-based global user task listeners for Self-Managed deployments.
Administrators can define cluster-wide listeners using configuration files or environment variables, ensuring they are applied consistently from cluster startup and preserved across backup and restore operations.
All user task lifecycle events emit payloads containing full variable context and metadata, enabling standardized integrations across all processes.
Modeler
Desktop Modeler: Manage Camunda connections
You can now manage Camunda connections directly in Desktop Modeler:
- Add, edit, delete, and save multiple connections.
- Securely store credentials and connection settings.
- Deploy directly to saved connections.
- Select an existing Orchestration Cluster or add a new one during deployment.
This streamlines the deployment workflow and reduces setup friction.
Web Modeler: Create event templates
You can now create, discover, and apply templates for more BPMN event types, including message, signal, and timer, directly within the element template editor.
You can also create global event templates that:
- Are reusable across projects.
- Standardize event configurations (for example, message names or payload structures).
- Help ensure consistency across teams and models.
Web Modeler: Invite users via email
As a Self-Managed administrator, you can now invite users to Web Modeler projects via email across all OIDC providers, eliminating the need to wait for users to log in first.
- Email-based invitations work for all OIDC providers (Keycloak, Entra ID, Okta, Auth0), matching SaaS behavior.
- Keycloak no longer receives special treatment; all providers follow the same invitation workflow.
This enables faster project provisioning and a consistent administrator experience across identity providers.
Orchestration Cluster
Manage configuration with cluster variables
Camunda 8.9 now supports cluster variables, letting you centrally manage configuration across your cluster. You can access these variables directly in the Modeler using FEEL expressions:
| Variable | Scope | Priority |
|---|---|---|
camunda.vars.cluster | Global | Lowest |
camunda.vars.tenant | Tenant | Medium |
camunda.vars.env | Merged view with automatic priority | Highest |
For example, if the same variable exists in multiple scopes, the priority is as follows:
- A Tenant variable overrides a Global variable.
- A Process-level variable has the highest priority, overriding both.
This hierarchy allows you to create cascading configurations, where specific contexts override broader defaults.
Cluster variables support simple key-value pairs and nested objects, which you can access with dot notation for complex structures. You can manage all cluster variables via the Orchestration Cluster API.
Use Amazon Aurora for secondary storage
Camunda 8.9 now supports Amazon Aurora as a secondary data store for orchestration clusters, in addition to existing options.
- Supports Aurora PostgreSQL (compatible with PostgreSQL 14–17).
- Designed for secure, high-performance, cloud-native deployments.
- Seamless integration with AWS features, including:
- IAM / IRSA authentication.
- High availability and failover.
- Alignment with DBA best practices.
Helm charts and manual installation guides now include tested configurations and step-by-step references for Aurora, reducing operational complexity and accelerating adoption for AWS-centric organizations.
Process instance migration
Camunda 8.9 now supports migration of process instances that include ad-hoc subprocesses, covering both single-instance and multi-instance (parallel and sequential) variants.
With this enhancement, you can:
- Safely migrate running instances.
- Update AI agent flows.
- Modernize process definitions without losing execution state.
This unlocks more flexible, agent-driven orchestration and faster iteration on live automation.
RDBMS secondary storage
Camunda 8.9 Helm charts now support RDBMS as fully integrated secondary storage options for orchestration clusters, providing a first-class alternative to Elasticsearch and OpenSearch.
With this update, administrators can:
- Use RDBMS as an alternative to Elasticsearch or OpenSearch.
- Configure database connections directly in
values.yaml. - Enable advanced authentication and custom JDBC drivers.
This allows enterprises to run Camunda 8 on familiar, enterprise-managed RDBMS infrastructure aligned with existing security, backup, and compliance requirements.
8.9.0-alpha2
| Release date | Changelog(s) | Blog |
|---|---|---|
| 09 December 2025 | - |
Agentic orchestration
A2A Client connectors
Agent-to-Agent (A2A) Client connectors allow you to interact with remote agents using the A2A protocol.
| Connector | Description |
| A2A Client connector | Interact with A2A agents, by retrieving the remote agent’s Agent Card and sending messages to the agent. |
| A2A Client Polling connector | Poll for responses from asynchronous A2A tasks. Typically paired with the A2A Client connector when using the Polling response retrieval method. |
| A2A Client Webhook connector | Receive callbacks from remote A2A agents via HTTP webhooks. Typically paired with the A2A Client connector when using the Notification response retrieval method. |
These connectors support multi-agent collaboration scenarios when combined with the AI Agent connector, as well as providing the ability to discover remote agents, send messages, and receive responses through multiple mechanisms.
MCP client authentication and transport protocol
The Camunda Model Context Protocol (MCP) client now supports OAuth, API key, and custom header–based authentication.
- System administrators can configure secure, policy-compliant access for Camunda AI agents.
- AI developers can discover and invoke enterprise MCP tools safely without exposing open endpoints.
MCP client connectors now also support connections using the streamable HTTP transport protocol.
This feature introduces breaking changes in the element templates and the runtime configuration of the MCP Client. To learn more, see announcements.
Connectors
Amazon Textract connector improvements
The Amazon Textract connector is improved with input field visibility and polling fixes, new sections for enhanced usability, and updated documentation.
Azure Blob Storage connector supports OAuth 2.0
The Azure Blob Storage connector now supports OAuth2.0 authentication with Microsoft Azure.
Azure Blob Storage connector OAuth 2.0
Email connector supports SMTP no authentication mode
The Email connector now supports noAuth authentication mode for SMTP. This feature is useful for customers running local mail servers without authentication requirements.
Runtime performance improvements with virtual threads executor (Self-Managed)
Connectors now use a virtual threads executor by default, using Project Loom to improve performance and scalability.
This allows the connector runtime to handle a larger number of concurrent jobs with lower resource consumption, particularly benefiting I/O-bound workloads typical in connector operations.
Console
Bulk import secrets (SaaS)
You can now add/import secrets in Console by directly uploading or pasting the contents of a .env file.
- Key–value pairs are automatically parsed, validated, and added as secrets.
- This helps reduce configuration errors and copy-pasting when adding secrets.
Cluster description (SaaS)
You can now add a cluster description when creating a cluster or by editing the cluster settings. This helps you document context, ownership, or add operational notes without changing the cluster name.
Import cluster secrets (SaaS)
You can now import and export connector secrets between clusters within your organization.
Export a cluster’s secrets to a key-value file for backup or external workflows, and import secrets from another cluster in a single action. Imports automatically match keys, update existing values, create missing ones, and provide clear feedback on the result. Permissions are enforced so that only authorized users can perform these actions.
Usage metrics for licence model and tenant (Self-Managed)
Self-Managed environment usage metrics now support per-tenant reporting and align with Camunda’s updated licensing model based on the number of tenants.
This feature is already available in the Camunda 8.8 release for Camunda 8 SaaS.
Database and data storage
Configure external RDBMS in Helm
Configure an external relational database (RDBMS) as secondary storage for the Orchestration Cluster when deploying with Helm.
- Supports all databases listed in the RDBMS support policy.
- Includes full configuration parameters, history-cleanup options, and exporter settings.
- Describes how to load JDBC drivers via init containers, custom images, or mounted volumes.
- Provides steps to verify database connectivity.
Open-source OpenSearch support
You can now use the open-source OpenSearch project for data storage in a Self-Managed deployment.
- This allows you to run a fully open source observability stack without using Elasticsearch or the Amazon OpenSearch Service.
- For configuration instructions, see the updated Helm chart values and compatibility matrix.
RDBMS version support policy
A new Camunda 8 Relational Database Management System RDBMS support policy provides information about:
- Officially supported database versions.
- The process for adopting new database versions.
- Timelines for phasing out older database versions.
SQL and Liquibase database scripts
SQL and Liquibase scripts are provided for all Camunda-supported databases.
- These scripts include database and schema creation, drop, and upgrade routines.
- Scripts follow best practices for each supported database type and version.
- The full script package is distributed as part of the official Camunda distribution, available via GitHub or Artifactory.
Modeler
Element template signal support
Element templates now support reusable BPMN signals.
- The
bpmn:Signal#propertybinding allows you to set the name of abpmn:Signalreferred to by the templated element. - This binding is only valid for templates of events with
bpmn:SignalEventDefinition.
Element template bpmn:Signal binding
Web Modeler: Embedded web server changed from Undertow to Tomcat (Self-Managed)
Web Modeler now uses Apache Tomcat as an embedded web server instead of Undertow. Aligning Web Modeler logging with the Orchestration Cluster makes it easier for administrators to configure and maintain Self-Managed deployments.
Web Modeler: IP egress monitoring (SaaS)
A new /meta/ip-ranges REST API endpoint allows you to monitor SaaS Web Modeler egress IP addresses.
- For example, the endpoint is available at https://api.cloud.camunda.io/meta/ip-ranges.
- Send a GET request to the endpoint to retrieve a list of egress IP addresses.
- Only IP addresses for the related services are exposed (Web Modeler).
- You should periodically monitor this list via the API, and make any changes in your systems as required.
- Although expected changes are published via the API at least 24 hours in advance, in exceptional cases Camunda might have to update these addresses within 24 hours and without prior notice. See static outbound IP addresses.
8.9.0-alpha1
| Release date | Changelog(s) | Blog |
|---|---|---|
| 13 November 2025 | - |
JDBC driver management for RDBMS integrations
Camunda 8.9 introduces a standardized approach to JDBC driver management for RDBMS integrations in manual installations.
- A new
/driver-libdirectory separates Camunda-bundled drivers from customer-supplied ones, providing a clear and compliant structure for database connectivity. - Drivers that Camunda can legally distribute are included by default. Customers can add and configure their own drivers (for example, Oracle JDBC).
- Configuration options allow full control, including explicit driver-class designation when required.
This change simplifies compliance and setup for RDBMS environments, ensuring consistent connectivity across PostgreSQL, Oracle, MariaDB, and H2.
MySQL and Microsoft SQL Server secondary storage
Camunda 8.9 extends RDBMS secondary storage to include MySQL and Microsoft SQL Server as additional database options for the Orchestration cluster.
- This enhancement provides greater flexibility for enterprises that depend on these databases due to policy, licensing, or ecosystem requirements, enabling smoother onboarding and infrastructure alignment.
- Zeebe’s primary execution storage remains Raft + RocksDB.
This alpha release introduces foundational support only. External configuration and Operate integration follows in upcoming alpha releases.
RDBMS secondary storage (H2, PostgreSQL, Oracle, MariaDB)
Camunda 8.9 introduces RDBMS secondary storage as an alternative to Elasticsearch or OpenSearch for storing and querying process data.
This feature enables organizations to use relational databases such as H2, PostgreSQL, Oracle, or MariaDB as the secondary storage layer, reducing operational complexity for teams that do not need the scale or performance of Elasticsearch or OpenSearch and prefer an RDBMS-based solution.
Key highlights:
- Flexible database choice: Use relational databases instead of Elasticsearch or OpenSearch.
- Separation of concerns: Zeebe’s primary execution storage remains Raft + RocksDB; this update only extends the secondary storage layer.
- Consistent APIs: Continue using the same REST API and data format as with Elasticsearch or OpenSearch—no query or integration changes needed.
- Simplified operations: Leverage existing RDBMS expertise without maintaining Elasticsearch or OpenSearch clusters.
This alpha release introduces support for H2 in Camunda 8 Run only. Operate and external RDBMS configuration follows in upcoming alpha releases.
Web Modeler: RDBMS support (H2, MariaDB, MySQL)
Web Modeler now supports H2, MariaDB, and MySQL as relational database systems, aligning with the configurations supported by the Orchestration cluster.
This enhancement ensures consistency across environments, simplifies setup for administrators, and improves integration for both SaaS and Self-Managed deployments.