Agents that get out of your way.

One line to start. Any model. Tools from a type hint. Python and TypeScript, same API.

pip install nobelagentic npm i nobelagentic Read the docs
from nobelagentic import Agent, tool

@tool
def get_weather(city: str) -> str:
    """Current weather for a city."""
    ...

agent = Agent("You help plan trips",
              tools=[get_weather, "web"])
agent.run("Plan a 3-day trip to Lagos")

Complexity is opt-in, never required.

Most frameworks make you learn their graph, their crew, or their chain on day one. Here you start with one agent and reach for more structure only when you need it. All three levels run on the same kernel, so streaming, tracing and checkpoints behave identically everywhere.

researcher = Agent(
    "Find and summarize sources",
    tools=["web", "files"],
    mcp=["github"],
)
researcher.run("What changed in Python 3.14?")

Tools from a type hint. Packs by name.

A function with type hints and a docstring is a tool. Built-in packs load lazily, declare what credentials they need, and stay out of the context window until you ask for them.

web

Search, fetch and clean pages, raw HTTP requests.

files

Read, write, list, regex-search — sandboxed to a root.

shell

Commands and in-process code. Opt-in only.

time

Now in any timezone, date math, conversions.

math

Safe expression evaluation and unit conversion.

data

JSON paths, CSV in and out, quick column stats.

Any MCP server is a tool pack.

Name a server and its tools appear next to your own, indistinguishable from native ones. A built-in registry covers GitHub, Postgres, Playwright, the filesystem and more; add your own with one command.

Agent(mcp=["github", "postgres"])

# or a remote server
Agent(mcp=["https://tools.example.com/mcp"])

# register one for every project
$ nobelagentic mcp add weather "npx -y weather-mcp"

One provider interface. Every model behind it.

Set an API key and the model is picked for you. Name a model and the provider is inferred from it. Swap vendors without touching your agent code.

AnthropicOpenAIGeminiOllamaGroqOpenRouterTogetherBring your own

Everything an agent needs to ship.

What's here today works today. What's coming has a version on it — no vaporware, the roadmap is public.

Streaming events

Token deltas, tool calls and usage as one typed event stream — the same in Python and TypeScript.

Handoffs & orchestration

Transfer the conversation to a specialist — or keep control: worker.as_tool() lets a supervisor delegate and combine.

Parallel tool calls

When the model asks for three tools at once, they run at once.

Structured outputs v0.2

Get a validated Pydantic model or zod object back, not a string to parse.

Sessions v0.2

Multi-turn conversations with pluggable storage. No hand-rolled history.

Human-in-the-loop v0.2

Checkpoint before a risky tool runs, resume after a human approves.

Memory & knowledge v0.3

Vector memory, plus retrieval over your own docs with one argument.

Observability v0.3

OpenTelemetry traces and per-run cost tracking. See exactly what your agents are doing.

Evals v0.4

Record a real run once, replay it offline in CI — assert on tool calls, outputs and cost.

Agents people actually ship.

The Examples page walks through each of these — complete, runnable, webhook included.

WhatsApp support agent

WhatsApp Cloud API webhook → agent with an order-lookup tool → reply. One file with FastAPI.

Read the recipe →

A support team

Billing and tech specialists behind a triage agent that routes with handoffs and never answers itself.

Read the recipe →

Telegram bot in TypeScript

Long-polling bot with a live FX-rate tool. No bot framework, ~25 lines.

Read the recipe →

How it compares

NobelAgenticOpenAI Agents SDKCrewAILangGraph
Lines to a working agent with a tool~6~12~25~40
Python and TypeScript, same APIYesYesPythonYes
Any provider without an adapter packageYesOpenAI-firstVia LiteLLMVia LangChain
Built-in tool packs20+, growingFewManyVia LangChain
MCP clientBy nameYesYesAdapter
Graph, crew and single-agent in one runtimeYesAgent + handoffsCrew + flowsGraph

Line counts are from each project's own quickstart as of October 2026. Crew and Graph APIs are on the roadmap for v0.2–0.3.

Questions people ask

What is NobelAgentic?

An open-source (MIT) agent framework for Python and TypeScript. One kernel with three API levels — Agent, Crew, Graph — so you start with one line and add structure only when you need it.

Which models and providers does it support?

Anthropic, OpenAI, Gemini, Ollama, Groq, OpenRouter, Together — any OpenAI-compatible host — or bring your own provider class. Set an API key and a model is picked for you; name a model and the provider is inferred.

Does it support MCP?

Yes, as a client: name a registry server (mcp=["github"]), point at a remote URL, or spawn a local one. MCP server mode — your agents as tools for any MCP client — lands in v0.4.

How do agents keep memory and context?

Today you prepend recent history (the Examples page shows the pattern). Sessions with pluggable storage land in v0.2; vector memory and knowledge sources in v0.3.

Python or TypeScript?

Both, with the same API and the same event stream — pip install nobelagentic or npm i nobelagentic. The two packages are ports of one shared spec.

How is it different from CrewAI, LangGraph or Mastra?

Those frameworks make you learn their abstraction on day one. Here complexity is opt-in: a working agent with a tool is ~6 lines, and crews and graphs compile to the same runtime when you grow into them — streaming, tracing and checkpoints behave identically at every level.

Is it production ready?

v0.1 is young and the API may still move before v1. What exists is tested (offline, with a scripted fake provider) and the roadmap is deliberately boring-first: structured outputs, sessions, run limits and human-in-the-loop are next.

Start building today.

An agent with a real tool, in the next five minutes.

pip install nobelagentic npm i nobelagentic Read the docs See examples