Practical overview
When it fits
Use DriftKit when building LLM agents or workflows in Java and when prompt lifecycle management should sit alongside that development. Choose its Context Engineering starter when a Spring Boot application and MongoDB-backed prompt-development records fit the project.
Who it is for
- Java developers building LLM agents or AI workflows.
- Teams that want prompt versions, tests and traces as part of agent development.
- Spring Boot developers who want a prompt engineering UI backed by MongoDB.
Documented uses
- Define a Java LLM agent with a configured model client and system message.
- Coordinate multi-agent workflows in a Java application.
- Manage prompt versions, tests and traces during development.
- Use the Context Engineering platform to work with prompts, traces, test sets and evaluations.
- Add audio processing to assistant or chatbot applications.
Requirements & limits
The project is open source under the Apache-2.0 license. Modules are published to Maven Central under the ai.driftkit group.
- Java 21 and Maven are required; starter modules require Spring Boot 3.3.x.
- The agent and workflow-engine modules can be used from plain Java without starting a Spring context.
- The Context Engineering platform uses MongoDB for traces, test sets, evaluation runs, audit logs and environments; its prompt store alone can use MongoDB, filesystem or in-memory storage.
- The platform UI's login form is not built-in authentication; production deployments need a separately configured security layer.
- A model provider API key and provider-accepted model ID are part of the documented agent setup.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up a plain Java LLM agent
Add the documented agent and OpenAI client dependencies, configure a model client, and define an agent.
Before you start
- Java 21
- Maven 3.8 or later is known to work; no Maven wrapper is included.
- Dependencies from the ai.driftkit Maven Central group.
- An API key for the chosen model provider and a provider-accepted model ID.
Steps
Add the driftkit-workflow-engine-agents and driftkit-clients-openai dependencies to the Maven project.
Create a VaultConfig with the provider name, API key and model ID.
Create a model client with ModelClientFactory.fromConfig(config).
Build an LLMAgent with the model client and a system message.
Expected outcome
A Java application has the documented dependencies and agent configuration in place; the README's example then calls executeText with a text prompt.
Guide limitations
- This recipe describes the setup shown in the README and does not establish that it has been independently executed.
- The example uses OpenAI; the README also names Gemini, Claude and DeepSeek provider IDs.
Configure the Context Engineering Spring Boot starter
Set up the documented starter configuration for the prompt engineering UI and its MongoDB-backed records.
Before you start
- Java 21 and Maven.
- Spring Boot 3.3.x for starter modules.
- MongoDB for traces, test sets, evaluation runs, audit logs and environments.
- A model provider API key and model ID.
Steps
Add the driftkit-context-engineering-spring-boot-starter and a model-client dependency to the Maven project.
Configure the MongoDB URI and provider credentials in application.yml.
Set driftkit.promptService.name to mongodb, filesystem or in-memory; this setting is mandatory.
Start MongoDB and run the Spring Boot application.
Expected outcome
The configured application provides the prompt engineering UI at http://localhost:8080/prompt-engineering, with prompts, traces, test sets and evaluations available through the UI and documented REST API.
Guide limitations
- MongoDB is required for the platform's traces, test sets, evaluations, audit log and environments; only the prompt store can use filesystem or in-memory storage.
- The UI login form is not built-in authentication. The README advises placing the application behind an independently configured security layer in production.
Compare options for Agent orchestration
These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.
| Project | Documented purpose | Platform | Access and requirements |
|---|---|---|---|
| Driftkit Framework | Define a Java LLM agent with a configured model client and system message. | CLI | The project is open source under the Apache-2.0 license. Modules are published to Maven Central under the ai.driftkit group. |
| Agenite | Define typed tools with JSON Schema validation and error handling. | CLI | |
| Veldra | Create an agent with a policy, tools, skills, knowledge base and reasoning method. | Self-hosted | The documented deployment is self-hosted and uses a sign-in gate after a first-run administrator account is created. |
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- driftkit-ai/driftkit-framework · Official homepage title · retained in its source language