DocsBlogSpecAbout
fabric ·nu.cc

Claude Code, as a Ref.

Declare a PromptRef on a Service, bind it with options once, then prompt Claude Code from inside your Nu tree.

Powered byclaude-agent-sdk
Read the referenceSee the code

See it.

Ask Claude Code anything, right from your Nu program. Wrap a few asks in a Session and it remembers what you said before.

Prompts from inside your Nu tree.

PromptRef.method() declares the endpoint; nu.cc.bind(Agent, ...) hands it a ClaudeAgentOptions template — model, cwd, tools, system prompt, permissions.

nu.cc.Session(...) is a bracket. The first prompt starts a fresh cc session, the rest continue it via resume=session_id. Sibling brackets stay independent.

agent.py
python · 14 locpy
import nu
class Agent(nu.Service):
ask = nu.cc.PromptRef.method()
# two independent sessions, two prompts each
app = nu.With(
nu.cc.bind(Agent, permission_mode="bypassPermissions"),
body=nu.cc.Session(
Agent.ask(prompt="my favorite color is teal. remember it."),
Agent.ask(prompt="what color did i tell you? one word."),
),
)
await nu.arun(app)

What you can do with it.

Everything cc can do — tools, agent loops, edits, custom system prompts — reachable from a plain Ref call.

options template

Bind once, prompt many times.

Model, cwd, allowed_tools, system_prompt, permission_mode, max_turns — set on bind, override per call when you need to.

sessions

Scope multi-turn work with a bracket.

nu.cc.Session wraps a subtree. Prompts inside share one cc session so context carries. Two sibling brackets = two clean contexts.

text + meta

Every call returns cost, turns, session id.

Prompt yields a dict: text, session_id, total_cost_usd, duration_ms, num_turns. Log it, cache it, put it in a UI — like any Ref result.

Combines well with.

nu.cc drives cc. Other fabrics hold the surrounding context and put the answers to work.

nu.kv

Durable state fabric.

Persist prompts, session ids, transcripts, tool traces. Resume conversations across process restarts.

nu.llm

OpenAI-compatible chat.

Same shape, different backend. Use nu.llm for plain chat/completions; use nu.cc when you want cc's agent loop and tools.

Try Nu.

One command gets you the wheel with every fabric. Then follow the movies tutorial to build a real app in an afternoon.

01 Install

pip install "nustack-py[all]"

02 Run the demo

nu demo movies

03 Build your app

Browse examples

Like what you see?

The project is young. Star it, join the room, watch what we ship next.

Star on GitHubnustackdev/nuJoin DiscordTalk to the team, share what you build.Follow updatesShip notes and small demos on X.