Skip to content

Installation

Requirements

  • Java 21 or later
  • Gradle or Maven build tool
  • A LangChain4j-supported LLM provider (OpenAI, Anthropic, Ollama, etc.)

Add the Dependency

agentensemble-bom is a Bill of Materials that aligns every AgentEnsemble module to the same version. Import it once and add only the modules you need -- no individual version numbers required.

Gradle (Kotlin DSL):

repositories {
    mavenCentral()
}

dependencies {
    // Import the BOM -- aligns all AgentEnsemble module versions
    implementation(platform("net.agentensemble:agentensemble-bom:{{ae_version}}"))

    // Core framework -- always required
    implementation("net.agentensemble:agentensemble-core")

    // Optional: task-scoped cross-execution memory (MemoryStore SPI)
    implementation("net.agentensemble:agentensemble-memory")

    // Optional: human-in-the-loop review gates (ReviewHandler SPI, ConsoleReviewHandler)
    implementation("net.agentensemble:agentensemble-review")

    // Optional: built-in tools -- add only the ones you need
    implementation("net.agentensemble:agentensemble-tools-web-search")
    implementation("net.agentensemble:agentensemble-tools-web-scraper")
    implementation("net.agentensemble:agentensemble-tools-calculator")
    implementation("net.agentensemble:agentensemble-tools-datetime")
    // Other tools: agentensemble-tools-json-parser, agentensemble-tools-file-read,
    // agentensemble-tools-file-write, agentensemble-tools-process, agentensemble-tools-http

    // Optional: live execution dashboard -- embedded WebSocket server that streams
    // real-time task/tool events to a browser and supports browser-based review gates.
    // agentensemble-review is included transitively; declare it above only if you
    // reference review types directly in your own code.
    implementation("net.agentensemble:agentensemble-web")

    // Optional: coding agent factory (auto-detects project, assembles coding tools)
    implementation("net.agentensemble:agentensemble-coding")

    // Optional: Java coding tools (GlobTool, CodeEditTool, ShellTool, etc.)
    implementation("net.agentensemble:agentensemble-tools-coding")

    // Optional: MCP bridge (filesystem + git via MCP servers; requires Node.js)
    implementation("net.agentensemble:agentensemble-mcp")

    // Optional: Micrometer metrics integration
    implementation("net.agentensemble:agentensemble-metrics-micrometer")

    // Add the LangChain4j integration for your LLM provider:
    implementation("dev.langchain4j:langchain4j-open-ai:1.11.0")
}

Gradle (Groovy DSL):

repositories {
    mavenCentral()
}

dependencies {
    implementation platform('net.agentensemble:agentensemble-bom:{{ae_version}}')
    implementation 'net.agentensemble:agentensemble-core'
    implementation 'net.agentensemble:agentensemble-memory'
    implementation 'net.agentensemble:agentensemble-review'
    implementation 'dev.langchain4j:langchain4j-open-ai:1.11.0'
}

Maven:

<dependencyManagement>
    <dependencies>
        <dependency>
            <groupId>net.agentensemble</groupId>
            <artifactId>agentensemble-bom</artifactId>
            <version>{{ae_version}}</version>
            <type>pom</type>
            <scope>import</scope>
        </dependency>
    </dependencies>
</dependencyManagement>
<dependencies>
    <dependency>
        <groupId>net.agentensemble</groupId>
        <artifactId>agentensemble-core</artifactId>
        <!-- version from BOM -->
    </dependency>
    <dependency>
        <groupId>net.agentensemble</groupId>
        <artifactId>agentensemble-memory</artifactId>
        <!-- version from BOM -->
    </dependency>
    <dependency>
        <groupId>net.agentensemble</groupId>
        <artifactId>agentensemble-review</artifactId>
        <!-- version from BOM -->
    </dependency>
    <dependency>
        <groupId>dev.langchain4j</groupId>
        <artifactId>langchain4j-open-ai</artifactId>
        <version>1.11.0</version>
    </dependency>
</dependencies>

Minimal: Core Only

If you need only the framework core (no memory, no review, no built-in tools):

dependencies {
    implementation("net.agentensemble:agentensemble-core:{{ae_version}}")
    implementation("dev.langchain4j:langchain4j-open-ai:1.11.0")
}

Tools-Only BOM

If you want to align just the built-in tool versions without importing the full BOM:

dependencies {
    implementation(platform("net.agentensemble:agentensemble-tools-bom:{{ae_version}}"))
    implementation("net.agentensemble:agentensemble-core:{{ae_version}}")

    // No version needed for tools -- resolved from tools BOM
    implementation("net.agentensemble:agentensemble-tools-calculator")
    implementation("net.agentensemble:agentensemble-tools-web-search")
}

Available Modules

Module Description
agentensemble-core Framework core -- required
agentensemble-tools-calculator Arithmetic expression evaluator
agentensemble-tools-datetime Date/time operations
agentensemble-tools-json-parser JSON path extraction
agentensemble-tools-file-read Sandboxed file reading
agentensemble-tools-file-write Sandboxed file writing
agentensemble-tools-web-search Web search (Tavily/SerpAPI)
agentensemble-tools-web-scraper Web page text extraction
agentensemble-tools-process Subprocess execution (cross-language)
agentensemble-tools-http HTTP endpoint wrapping
agentensemble-tools-coding Coding tools: glob, code search, code edit, shell, git, build runner, test runner
agentensemble-tools-bom Version alignment BOM
agentensemble-coding Coding agent factory: auto-detects project type, assembles tools, generates coding prompts
agentensemble-mcp MCP protocol bridge: adapts MCP server tools to the AgentTool interface
agentensemble-workspace Workspace isolation: git worktree management for coding agents
agentensemble-memory Memory subsystem: task-scoped cross-execution memory with MemoryStore SPI; also includes legacy short-term, long-term, and entity memory
agentensemble-review Human-in-the-loop review gates: ReviewHandler SPI, ConsoleReviewHandler, and HumanInputTool
agentensemble-web Live execution dashboard: embedded WebSocket server that streams real-time task/tool/delegation events to a browser and supports browser-based review gates
agentensemble-metrics-micrometer Micrometer metrics integration
agentensemble-bom Top-level BOM -- aligns all module versions in one import

Supported LLM Providers

AgentEnsemble uses the LangChain4j ChatModel interface and works with any provider that implements it:

Provider LangChain4j Artifact
OpenAI (GPT-4o, GPT-4o-mini, etc.) langchain4j-open-ai
Anthropic (Claude 3.5, Claude 3 Haiku, etc.) langchain4j-anthropic
Ollama (local models) langchain4j-ollama
Azure OpenAI langchain4j-azure-open-ai
Amazon Bedrock langchain4j-bedrock
Google Vertex AI langchain4j-vertex-ai-gemini
Mistral AI langchain4j-mistral-ai

For a full list, see the LangChain4j documentation.

Logging

AgentEnsemble uses SLF4J for logging. Add your preferred implementation:

// Logback (recommended)
implementation("ch.qos.logback:logback-classic:1.5.32")

See the Logging guide for configuration details.

Execution Graph Visualizer: agentensemble-viz

agentensemble-viz is a standalone developer tool for visualizing task dependency graphs and execution timelines. It reads JSON files exported by agentensemble-devtools and renders them in a local web UI.

Homebrew (macOS and Linux):

brew install agentensemble/tap/agentensemble-viz
agentensemble-viz ./traces/

This installs a self-contained native binary (no Node.js required). The formula is updated automatically on every AgentEnsemble release.

npx (no installation required):

npx @agentensemble/viz ./traces/

Global npm install:

npm install -g @agentensemble/viz
agentensemble-viz ./traces/

See the Visualization Guide for the complete workflow including how to export DAGs and execution traces from your Java application.

Next Steps