langchain-assistanthub

Crypto intelligence toolkit for LangChain agents. Real-time prices, risk scores, Monte Carlo backtests, AI forecasts, and more — powered by Assistant Hub.

Install

# Recommended: pin to current stable version
pip install langchain-assistanthub==0.1.4

# Or latest (for bleeding-edge users)
# pip install langchain-assistanthub

We're at v0.1.4 — pinned install recommended for stability. Check the changelog for updates!

Optional extras:

pip install langchain-assistanthub[x402]      # USDC on-chain payments
pip install langchain-assistanthub[swarm]     # LangGraph multi-agent
pip install langchain-assistanthub[notebook]  # JupyterLab + pandas + matplotlib for examples
pip install langchain-assistanthub[dev]       # pytest, ruff, mypy

Quick Start

Option A: Pre-built toolkit (recommended)

from langchain_assistanthub import AssistantHubToolkit
from langgraph.prebuilt import create_react_agent

toolkit = AssistantHubToolkit.from_api_key("ahk_your_key")
# or: toolkit = AssistantHubToolkit.from_env()   # reads ASSISTANT_HUB_API_KEY
tools = toolkit.get_tools()

agent = create_react_agent(model, tools)
result = agent.invoke({"messages": [
    {"role": "user", "content": "What's the risk score for SOL?"}
]})

Option B: Login with Hub credentials

from langchain_assistanthub import AssistantHubToolkit

toolkit = AssistantHubToolkit.from_hub_login("you@email.com", "password")
tools = toolkit.get_tools()

Option C: MCP auto-discovery

Dynamically discover tools from the MCP server — picks up new tools without upgrading the package:

from langchain_assistanthub import AssistantHubToolkit

tools = await AssistantHubToolkit.from_mcp(api_key="ahk_your_key")
agent = create_react_agent(model, tools)

Option D: MCP client directly

from langchain_assistanthub import AssistantHubMCPClient

client = AssistantHubMCPClient.from_api_key("ahk_your_key")
tools = await client.get_tools()

Available Tools

* Requires enable_price_feed=True and websockets package.

Advanced Usage

# Free tools only
toolkit = AssistantHubToolkit(api_key="...", include_premium=False)

# Specific tools
toolkit = AssistantHubToolkit(api_key="...", tools=["live_prices", "risk_scores"])

# Custom instance (local dev)
toolkit = AssistantHubToolkit(api_key="...", base_url="http://localhost:3000")

# With live price monitoring
toolkit = AssistantHubToolkit(
    api_key="...",
    enable_price_feed=True,
    price_feed_coins=["BTC", "ETH", "SOL"],
)

Anonymous Free Tier

No API key? You get 10 free calls per day on non-premium tools:

toolkit = AssistantHubToolkit()  # no api_key needed
tools = toolkit.get_tools()     # 10 calls/day, then 429 with upgrade CTA

Payment

Free tools work instantly. Premium tools require:

• Pro/Premium JWT — unlimited calls
• x402 USDC on Base — pay per call (no subscription needed)
• HUB stakers — 50% off all x402 calls

Package Structure

sdk/langchain-python/
  src/langchain_assistanthub/
    __init__.py          # Public API + __all__
    _version.py          # Single source of truth for version
    toolkit.py           # AssistantHubToolkit (main entry)
    client.py            # AssistantHubMCPClient (MCP adapter)
    tools.py             # Individual tool wrappers
    price_feed.py        # WebSocket price feed
    price_monitor.py     # Price alert monitor
    strategy.py          # Strategy analysis tool
    execution.py         # Trade execution tools
  tests/
    test_toolkit.py      # Comprehensive test suite
  pyproject.toml         # hatchling build config

Example Notebooks

See sdk/notebooks/ for runnable examples:

Telemetry

This package sends a single anonymous ping on toolkit init to help us gauge adoption. No PII is collected — just a random ID, package version, and auth type (keyed vs anonymous).

Opt out:

export ASSISTANT_HUB_TELEMETRY_OPT_OUT=1

Links

• Documentation
• MCP Discovery
• GitHub