Back to directory

ai4j

A Java 8+ agentic development toolkit that unifies access to more than a dozen large-language-model platforms behind one API, with built-in tool calling, RAG, MCP client and server support, and a long-running agent harness.

ai4j

ai4j provides one factory for chat, responses and messages-style protocols across OpenAI, Anthropic, DeepSeek, Zhipu, Doubao, Ollama and other platforms, with function calling and native streaming, plus embedding, reranking and media-generation service interfaces. Switching providers only requires changing a platform type and its configuration. Its agent layer supports ReAct, CodeAct and Deep Research execution modes with graph-based orchestration for conditional branches and loops, subagent delegation and multi-agent teams.

Its harness component turns a normally one-shot agent call into a pausable, resumable and auditable long-running task: state persists to file or JDBC storage, multiple workers claim tasks safely by lease, and a submitted result must pass a review and acceptance gate before it counts as complete, with accepted decisions able to be promoted into runtime rules. Built-in RAG covers document loading, chunking, eight vector-store backends, hybrid retrieval, reranking and citation, and it installs as a Spring Boot starter or a plain Maven or Gradle dependency.

More in Agent frameworks

See all 902
  • Recurse

    recurse.run

    A serverless harness that lets a coding agent design, test, and deploy its own specialist sub-agents as tools, MCP servers, or bots, defined by a schema-validated manifest.

  • ProofForge

    shulgin.is-a.dev

    An AI-agent pipeline that decomposes a mathematics problem, proves each piece, and formalizes the result in Lean 4 and Mathlib until the Lean kernel accepts it, with six of its proofs already merged into Google DeepMind's formal-conjectures repository.

  • local-coder

    github.com

    Builds a local coding-agent team on top of Ollama and OpenCode by detecting the machine's hardware, recommending models for seven specialist roles, and testing whether each assigned model can actually use tools, edit a repository and pass verification.