01MODEL
Unified LLM
OpenAI, Anthropic and other models share one interface, keeping upper layers free from vendor lock-in.
AgentForge v1.0 · Java Agent Framework
Intelligence that goes beyond answers
and takes real action.
From unified LLM abstractions to Agent Runtime, AgentForge gives Java developers a transparent, controllable and extensible foundation for building agents.
Java native0 vendor lock-inMIT open source
agent-run.traceLIVE
RUN / AF-104232ms
USER
Inspect the project and propose the next implementation plan.
- 01THINKdone
Identify the goal and current context
- 02TOOLdone
repository.inspect()
- 03ACTrunning
Generate an executable task plan
HOW IT WORKS
A clear path from intent to action
Every layer has a clear boundary. Every step can be observed, replaced and extended.
INPUTGoalNatural-language intent
→REASONThinkContext and planning
→EXECUTEToolInvoke real capabilities
→OUTPUTActionComplete the task loop
BUILDING BLOCKS
Built for agent engineering
02RUNTIME
Agent Runtime
A ReAct loop coordinates reasoning, action, observation and stopping in a transparent runtime.
03ACTION
Tool Calling
Typed tool definitions, argument parsing and execution results reliably turn model decisions into action.
04CONTEXT
Memory & Context
Combine conversations, runtime state and long-term memory into an extensible agent context.
READY TO FORGE?
Read the quick start↗