Skip to main content
Wikitopia
Get API Key

LangChain Integration

← Docs

Three patterns for using Wikitopia's knowledge graph in LangChain agents.

Pattern 1 — LangChain Tool (simplest)

Wrap the Wikitopia REST API as a LangChain Tool. Works with any LLM.

from langchain.tools import Tool
import requests

API_URL = "https://api.wikitopia.org"

def search_wikitopia(query: str) -> str:
    """Search for verified facts about AI tools, models, and companies."""
    res = requests.get(f"{API_URL}/v1/search",
        params={"q": query, "mode": "hybrid", "limit": 5})
    entities = res.json().get("results", [])
    return "\n".join([
        f"{e['canonical_name']} ({e['entity_type']}): "
        f"{e.get('description', 'No description')[:120]}"
        for e in entities
    ]) or "No results found."

def get_entity(name: str) -> str:
    """Get detailed info about a specific AI entity."""
    res = requests.get(f"{API_URL}/v1/entities/{name}")
    if res.status_code == 404:
        return f"Entity '{name}' not found."
    e = res.json()
    props = e.get("properties", {})
    lines = [f"# {e['canonical_name']} ({e['entity_type']})"]
    if e.get("description"):
        lines.append(e["description"])
    for key in ["developed_by", "license_type", "founded_year", "primary_use_case"]:
        if props.get(key):
            lines.append(f"{key}: {props[key]}")
    lines.append(f"Claims: {e.get('claim_count', 0)}")
    return "\n".join(lines)

wikitopia_search = Tool(
    name="WikitopiaSearch",
    description="Search Wikitopia for verified facts about AI tools, "
                "models, frameworks, and companies. Input: search query string.",
    func=search_wikitopia,
)

wikitopia_entity = Tool(
    name="WikitopiaEntity",
    description="Get detailed profile of a specific AI entity by name. "
                "Input: entity name like 'LangChain' or 'OpenAI'.",
    func=get_entity,
)

Usage with an agent

from langchain.agents import initialize_agent, AgentType
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4o-mini")
agent = initialize_agent(
    tools=[wikitopia_search, wikitopia_entity],
    llm=llm,
    agent=AgentType.OPENAI_FUNCTIONS,
    verbose=True,
)

result = agent.run("Compare the top 3 open-source vector databases")
print(result)

Pattern 2 — Community Package (coming soon)

# pip install langchain-wikitopia  (in development)
from langchain_wikitopia import WikitopiaSearchTool, WikitopiaEntityTool

tools = [
    WikitopiaSearchTool(),        # Wraps /v1/search
    WikitopiaEntityTool(),        # Wraps /v1/entities/:name
]

A dedicated langchain-wikitopia package is in development with built-in caching, type-safe outputs, and async support. Watch the repo for updates.

Pattern 3 — MCP via LangChain

Use the Wikitopia MCP server with Claude or any MCP-compatible model.

# Install the MCP server globally
npx wikitopia-mcp

# Or add to claude_desktop_config.json:
{
  "mcpServers": {
    "wikitopia": {
      "command": "npx",
      "args": ["-y", "wikitopia-mcp"]
    }
  }
}

The MCP server provides 8 tools including wikitopia_search, wikitopia_get_entity, wikitopia_compare, and wikitopia_explore_ecosystem.

See also