v0.1.0 · open source

Build your Next Agent.

NxAgent is a production-grade agentic framework for orchestrating multi-agent workflows, tool execution, and intelligent task routing.

$ pip install nx-agent click to copy
example.py
# Build a multi-agent research workflow in minutes
from nx_agent import Agent, Workflow, tool

# Define tools
@tool
def web_search(query: str) -> str:
    """Search the web for current information."""
    ... # your implementation

# Define agents
researcher = Agent(
    role="Research Analyst",
    goal="Find and synthesize accurate information",
    tools=[web_search],
)

writer = Agent(
    role="Technical Writer",
    goal="Produce clear, structured reports",
)

# Orchestrate as a workflow
workflow = Workflow(agents=[researcher, writer])
result = workflow.run("Summarize the latest advances in agentic AI")
print(result)
// features

Everything you need.
Nothing you don't.

Multi-agent orchestration

Define agent roles, assign tasks, and let NxAgent handle routing, delegation, and inter-agent communication automatically.

Tool ecosystem

Plug in web search, code execution, file I/O, and any custom API with a simple decorator-based tool SDK.

Memory & context

Short-term session memory, long-term vector store integration, and smart context window management built in.

LLM agnostic

Works with OpenAI, Anthropic, Gemini, Ollama, or any OpenAI-compatible endpoint. Swap providers in one line.

Streaming support

Real-time token streaming and async-first architecture so your agents never block your application.

Observability

Full trace logging, step-by-step agent reasoning, and structured output for every run. Debug what actually happened.

// workflow

From idea to
running agent.

01
Define your agents

Give each agent a role, goal, and backstory. NxAgent uses these to guide LLM behavior and keep agents on track across long-running tasks.

02
Attach tools

Decorate any Python function with @tool and it becomes available to your agents. NxAgent handles schema generation, calling, and error recovery automatically.

03
Orchestrate with a Workflow

Group agents into a Workflow and define how they collaborate — sequentially, in parallel, or with dynamic task routing based on agent expertise.

04
Run and observe

Execute your workflow and inspect every step — agent reasoning, tool calls, memory reads, and final outputs — with structured trace logging.

// applications

Built for real
agent workflows.

01
Research workflows

Coordinate agents that gather context, compare findings, verify claims, and produce concise summaries from multiple sources.

Market scans Research briefs Source synthesis
02
Tool-using assistants

Give agents access to Python functions, APIs, search, files, and internal systems through simple tool definitions.

API actions File operations Internal tools
03
Multi-agent production flows

Split planning, execution, review, and reporting across specialist agents that collaborate inside one workflow.

Task routing Review loops Traceable runs

Ship production agents faster.

Copied to clipboard!