> ## Documentation Index
> Fetch the complete documentation index at: https://lab.pollack.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Projects

> Tools for building, evaluating, and observing AI agents on the JVM

The lab produces a focused set of projects that work together as an agent engineering stack — for building, evaluating, and understanding AI agents on the JVM.

## Loopy

<Card title="Loopy" icon="terminal" href="/projects/loopy">
  Interactive coding agent CLI for Java developers — domain skills, modern TUI, Spring Boot scaffolding, multi-provider support. Where the thesis becomes a product.
</Card>

## AgentWorks

Core `agent-*` libraries that compose into a complete agent engineering stack. Use the [AgentWorks BOM](/projects/agentworks-bom) for coordinated version management.

<CardGroup cols={2}>
  <Card title="Agent Workflow" icon="arrows-spin" href="/projects/agent-workflow">
    Compose multi-step agentic pipelines — steps, typed context, branching, loops, quality gates, parallel execution. Compiles to a graph IR with portable runtimes.
  </Card>

  <Card title="Agent Client" icon="robot" href="/projects/agent-client">
    Autonomous CLI agent integrations — unified framework for invoking Claude Code, Gemini CLI, and SWE-bench agents as Spring services. Portable MCP servers, auto-configuration, sandbox isolation. Portable reasoning effort (`low`/`medium`/`high`) across providers in 0.21.0, with provider-native ranges (Claude `low`→`max`, Codex `minimal`→`xhigh`) when you need finer control.
  </Card>

  <Card title="Agent Judge" icon="gavel" href="/projects/agent-judge">
    Agent-agnostic evaluation with deterministic, command, and LLM judges. Jury system with voting strategies. Zero cost for T0-T1 checks — LLM fires only when cheaper tiers pass.
  </Card>

  <Card title="Agent Bench" icon="chart-bar" href="/projects/agent-bench">
    Benchmarking suite for Java-centric AI agents on enterprise dev tasks — issue triage, PR review, coverage, compliance. Agent-agnostic with provide/grade separation.
  </Card>

  <Card title="Agent Experiment" icon="flask" href="/projects/agent-experiment">
    End-to-end experiment driver — datasets, runner, jury scoring, variant comparison, sessions, and sweeps. The execution backbone of every lab experiment.
  </Card>

  <Card title="Agent Memory" icon="brain" href="/projects/agent-memory">
    Progressive memory management — context compaction, token-budgeted retrieval, LLM summarization. 21x token reduction in production. Spring AI BaseAdvisor integration.
  </Card>

  <Card title="Agent Hooks" icon="anchor" href="/projects/agent-hooks">
    Portable hook API for steering agent behavior at the tool-call boundary — deterministic safety, observability, and input modification that works across any runtime.
  </Card>

  <Card title="Agent Journal" icon="book" href="/projects/agent-journal">
    Behavioral trace capture — every tool call, state transition, and decision point. Feeds Markov fingerprinting and cross-variant behavioral comparison.
  </Card>

  <Card title="Agent Sandbox" icon="shield" href="/projects/agent-sandbox">
    Isolated command execution — local, Docker, or E2B cloud backends behind a unified API. Safe execution for untrusted agent code.
  </Card>

  <Card title="Agent Tools" icon="wrench" href="/projects/agent-tools">
    Claude Code-inspired tools for Spring AI agents — file I/O, shell, search, web, subagent orchestration.
  </Card>

  <Card title="Agent Skills" icon="book-sparkles" href="/projects/agent-skills">
    Curated domain knowledge modules (SkillsJars) — teach agents how to test, review, and build Spring applications.
  </Card>
</CardGroup>

## Agento

<CardGroup cols={2}>
  <Card title="Agento Studio" icon="wand-magic-sparkles" href="/projects/agento-studio">
    Visual environment for designing, composing, and deploying autonomous agents on the JVM. Coming soon.
  </Card>

  <Card title="Agento University" icon="graduation-cap" href="/projects/agento-university">
    Hierarchical multi-agent platform — campus stewards that continuously maintain software projects. Coming soon.
  </Card>
</CardGroup>

## SDKs & Protocols

Foundational SDKs for agent communication and CLI integration.

<CardGroup cols={2}>
  <Card title="Claude Agent SDK (Java)" icon="microchip" href="/projects/claude-agent-sdk">
    Java SDK for Claude Code CLI — three-API architecture, 23-part tutorial, Spring Boot auto-configuration
  </Card>

  <Card title="ACP Java SDK" icon="handshake" href="/projects/acp-java-sdk">
    Agent Communication Protocol — build agents, consume agents, integrate with Zed, JetBrains, and VS Code
  </Card>
</CardGroup>