metorial 2.3.0

Python SDK for Metorial - The open source integration platform for agentic AI

Metorial Python SDK

The official Python SDK for Metorial. Give your AI agents access to tools like Slack, GitHub, SAP, and hundreds more through MCP — without managing servers, auth flows, or infrastructure.

Sign up for a free account to get started.

Complete API Documentation

• Documentation - Documentation and guides
• API Reference - Complete API reference

Installation

pip install metorial

Supported LLM Integrations

This SDK formats MCP tools for each LLM provider. Pass the provider parameter to get tools in the right format.

Provider	Format	Client Library	Models (non-exhaustive)
OpenAI	provider="openai"	openai	gpt-4.1, gpt-4o, o1, o3
Anthropic	provider="anthropic"	anthropic	claude-sonnet-4-5, claude-opus-4
Google Gemini	provider="google"	google-generativeai	gemini-2.5-pro, gemini-2.5-flash
Mistral	provider="mistral"	mistralai	mistral-large-latest, codestral-latest
DeepSeek	provider="deepseek"	openai (compatible)	deepseek-chat, deepseek-reasoner
Together AI	provider="togetherai"	openai (compatible)	Llama-4, Qwen-3
xAI (Grok)	provider="xai"	openai (compatible)	grok-3, grok-3-mini

Framework Integrations

For popular agent frameworks, we provide helper functions that convert tools to the framework's native format:

Framework	Import	Example
PydanticAI	from metorial.integrations.pydantic_ai import create_pydantic_ai_tools	example
LangChain	from metorial.integrations.langchain import create_langchain_tools	example
LangGraph	from metorial.integrations.langgraph import create_langgraph_tools	example
OpenAI Agents	from metorial.integrations.openai_agents import create_openai_agent_tools	example
Haystack	from metorial.integrations.haystack import create_haystack_tools	example

Quick Start

This example uses PydanticAI with Anthropic Claude and Metorial Search, a built-in web search provider that requires no auth configuration. You just need two environment variables:

• METORIAL_API_KEY from platform.metorial.com
• ANTHROPIC_API_KEY from console.anthropic.com

pip install metorial pydantic-ai python-dotenv

import asyncio
import os

from metorial import Metorial, metorial_pydantic_ai
from pydantic_ai import Agent

metorial = Metorial(api_key=os.environ["METORIAL_API_KEY"])

async def main():
    deployment = metorial.provider_deployments.create(
        name="Metorial Search",
        provider_id="metorial-search",
    )

    session = await metorial.connect(
        adapter=metorial_pydantic_ai(),
        providers=[
            {"provider_deployment_id": deployment.id},
        ],
    )
    agent = Agent(
        "anthropic:claude-sonnet-4-20250514",
        system_prompt="You are a helpful research assistant.",
        tools=session.tools(),
    )

    result = await agent.run(
        "Search the web for the latest news about AI agents and summarize the top 3 stories."
    )
    output = getattr(result, "data", None) or getattr(result, "output", str(result))
    print(output)

asyncio.run(main())

See the full runnable example at examples/pydantic-ai/.

Authenticating MCP Tool Providers

The Quick Start above used Metorial Search, which requires no authentication. Most providers — Slack, GitHub, SAP, and others — require credentials. Here are the options, from simplest to most flexible.

Key concepts:

• Provider — an MCP tool integration (e.g. Slack, GitHub, Metorial Search). Browse available providers at platform.metorial.com.
• Provider Deployment — an instance of a provider configured for your project. You can create deployments in the dashboard or programmatically via metorial.provider_deployments.create().
• Auth Credentials — your OAuth app registration (client ID, client secret, scopes).
• Auth Config — an already-authenticated connection with a token, service account, or specific user via an OAuth flow.

Dashboard-Configured Deployments

Some providers (Exa, Tavily) use API keys configured entirely in the dashboard. Just pass the deployment ID — no auth code needed:

providers=[{"provider_deployment_id": "your-exa-deployment-id"}]

Pre-Created Auth Config

An auth config represents an already-authenticated connection to a provider — for example, a user who has completed the OAuth flow for Slack. Once created (via the dashboard or a setup session), reference it by ID:

providers=[
    {
        "provider_deployment_id": "your-slack-deployment-id",
        "provider_auth_config_id": "your-auth-config-id",
    }
]

Inline Credentials

Pass credentials directly without pre-creating them in the dashboard:

providers=[
    {
        "provider_deployment_id": "your-deployment-id",
        "provider_auth_config": {
            "provider_auth_method_id": "your-auth-method-id",
            "credentials": {"access_token": "user-access-token"},
        },
    }
]

OAuth Flow

For services like Slack or GitHub where each end-user authenticates individually, use setup sessions to handle the OAuth flow:

import os
from metorial import Metorial, metorial_pydantic_ai

metorial = Metorial(api_key=os.environ["METORIAL_API_KEY"])

# 1. Create a setup session for the provider
setup_session = metorial.provider_deployments.setup_sessions.create(
    provider_id="your-slack-provider-id",
    provider_auth_method_id="oauth",
    redirect_url="https://yourapp.com/oauth/callback",
)

# 2. Send the OAuth URL to your user
print(f"Authenticate here: {setup_session.url}")

# 3. Wait for the user to complete OAuth
completed = await metorial.wait_for_setup_session([setup_session])

# 4. Use the auth config in a connected session
session = await metorial.connect(
    adapter=metorial_pydantic_ai(),
    providers=[
        {
            "provider_deployment_id": "your-slack-deployment-id",
            "provider_auth_config_id": completed[0].auth_config.id,
        }
    ],
)
tools = session.tools()
# Use tools...

Multiple Providers in One Session

Combine providers freely in a single session — each can use a different auth method:

# Create a deployment for Metorial Search
deployment = metorial.provider_deployments.create(
    name="Metorial Search",
    provider_id="metorial-search",
)

providers=[
    # Metorial Search (no auth needed)
    {"provider_deployment_id": deployment.id},
    # Dashboard-configured deployment
    {"provider_deployment_id": "your-slack-deployment-id", "provider_auth_config_id": "slack-auth-config-id"},
    # Inline credentials
    {
        "provider_deployment_id": "your-github-deployment-id",
        "provider_auth_config": {
            "provider_auth_method_id": "github-auth-method-id",
            "credentials": {"access_token": "ghp_..."},
        },
    },
]

Session Templates

Pre-configure provider combinations on the dashboard, then reference them by ID. This is useful when you want to manage which providers and auth configs are used without changing code:

from metorial import metorial_pydantic_ai

# Reference a session template by ID
session = await metorial.connect(
    adapter=metorial_pydantic_ai(),
    providers=[
        {"session_template_id": "your-template-id"},
    ],
)
tools = session.tools()
# All providers from the template are available

# You can also mix session templates with explicit provider deployments
# in the same providers list
deployment = metorial.provider_deployments.create(
    name="Metorial Search",
    provider_id="metorial-search",
)

session = await metorial.connect(
    adapter=metorial_pydantic_ai(),
    providers=[
        {"session_template_id": "your-template-id"},
        {"provider_deployment_id": deployment.id},
    ],
)

Enterprise: Bring Your Own (BYO) Credentials

For enterprise deployments, you have flexible options:

• Shared deployment: Deploy once and share with all users (works well for API key-based tools like Exa, Tavily)
• BYO OAuth: For services like SAP, enterprises can register their own OAuth app credentials:

credentials = await metorial.provider_deployments.auth_credentials.create(
    provider_id="your-sap-provider-id",
    name="Our SAP OAuth App",
    config={
        "client_id": "your-client-id",
        "client_secret": "your-client-secret",
        "scopes": ["read", "write"],
    },
)

• Dynamic deployments: Create provider deployments programmatically via the Provider Deployment API.

Session Options

• Recommended entry point: Use metorial.connect(adapter=..., providers=[...]) for new code. It matches the Node SDK and does not require manual close calls.
• Compatibility wrapper: provider_session(...) still exists when you want the older provider-specific session helpers inside an async with, but it is now just a thin wrapper over the same adapter/session resolution path used by connect().
• Multiple providers: Pass multiple entries in the providers list to combine tools from different MCP servers.

Examples

Check out the examples/ directory for complete working examples:

Example	Framework	Description
pydantic-ai	PydanticAI + Anthropic	PydanticAI agent with tool calls
langchain	LangChain + Anthropic	LangChain agent with react pattern
langgraph	LangGraph + Anthropic	LangGraph streaming agent
openai-agents	OpenAI Agents SDK	OpenAI Agents with tool calls
haystack	Haystack + OpenAI	Haystack pipeline with tools

Provider Examples

These examples use connect() directly when you want provider-native tool formats rather than a framework adapter.

OpenAI

import os
from openai import AsyncOpenAI
from metorial import Metorial, metorial_openai

metorial = Metorial(api_key=os.environ["METORIAL_API_KEY"])
openai = AsyncOpenAI(api_key=os.environ["OPENAI_API_KEY"])

deployment = metorial.provider_deployments.create(
    name="Metorial Search",
    provider_id="metorial-search",
)

session = await metorial.connect(
    adapter=metorial_openai(),
    providers=[{"provider_deployment_id": deployment.id}],
)
messages = [{"role": "user", "content": "Search the web for the latest news about AI agents."}]

response = await openai.chat.completions.create(
    model="gpt-4o",
    messages=messages,
    tools=session.tools(),
)

if response.choices[0].message.tool_calls:
    results = await session.call_tools(response.choices[0].message.tool_calls)
    # Add results to messages and continue conversation...

Anthropic

import os
from anthropic import AsyncAnthropic
from metorial import Metorial, metorial_anthropic

metorial = Metorial(api_key=os.environ["METORIAL_API_KEY"])
anthropic = AsyncAnthropic(api_key=os.environ["ANTHROPIC_API_KEY"])

deployment = metorial.provider_deployments.create(
    name="Metorial Search",
    provider_id="metorial-search",
)

session = await metorial.connect(
    adapter=metorial_anthropic(),
    providers=[{"provider_deployment_id": deployment.id}],
)
response = await anthropic.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=1024,
    tools=session.tools(),
    messages=[{"role": "user", "content": "Search the web for the latest news about AI agents."}],
)

if response.stop_reason == "tool_use":
    tool_calls = [b for b in response.content if b.type == "tool_use"]
    results = await session.call_tools(tool_calls)
    # Add results to messages and continue conversation...

Google Gemini

import os
import google.generativeai as genai
from metorial import Metorial, metorial_google

metorial = Metorial(api_key=os.environ["METORIAL_API_KEY"])
genai.configure(api_key=os.environ["GOOGLE_API_KEY"])

deployment = metorial.provider_deployments.create(
    name="Metorial Search",
    provider_id="metorial-search",
)

session = await metorial.connect(
    adapter=metorial_google(),
    providers=[{"provider_deployment_id": deployment.id}],
)
model = genai.GenerativeModel("gemini-2.5-pro", tools=session.tools())
chat = model.start_chat()
response = chat.send_message("Search the web for the latest news about AI agents.")

for part in response.parts:
    if fn := part.function_call:
        result = await session.call_tool(fn.name, dict(fn.args))
        # Continue conversation with result...

Mistral

import os
from mistralai import Mistral
from metorial import Metorial, metorial_mistral

metorial = Metorial(api_key=os.environ["METORIAL_API_KEY"])
mistral = Mistral(api_key=os.environ["MISTRAL_API_KEY"])

deployment = metorial.provider_deployments.create(
    name="Metorial Search",
    provider_id="metorial-search",
)

session = await metorial.connect(
    adapter=metorial_mistral(),
    providers=[{"provider_deployment_id": deployment.id}],
)
response = await mistral.chat.complete_async(
    model="mistral-large-latest",
    tools=session.tools(),
    messages=[{"role": "user", "content": "Search the web for the latest news about AI agents."}],
)

if response.choices[0].message.tool_calls:
    results = await session.call_tools(response.choices[0].message.tool_calls)
    # Add results to messages and continue conversation...

OpenAI-compatible Models (DeepSeek, Together, xAI)

import os
from openai import AsyncOpenAI
from metorial import Metorial, metorial_openai_compatible

metorial = Metorial(api_key=os.environ["METORIAL_API_KEY"])
xai_compatible = AsyncOpenAI(
    api_key=os.environ["OPENAI_COMPATIBLE_API_KEY"],
    base_url="https://your-openai-compatible-endpoint/v1",
)

deployment = metorial.provider_deployments.create(
    name="Metorial Search",
    provider_id="metorial-search",
)

session = await metorial.connect(
    adapter=metorial_openai_compatible(),
    providers=[{"provider_deployment_id": deployment.id}],
)
response = await xai_compatible.chat.completions.create(
    model="your-model-name",
    tools=session.tools(),
    messages=[{"role": "user", "content": "Search the web for the latest news about AI agents."}],
)

if response.choices[0].message.tool_calls:
    results = await session.call_tools(response.choices[0].message.tool_calls)
    # Add results to messages and continue conversation...

Framework Integration Examples

LangChain / LangGraph

from metorial import metorial_langchain
from langchain_anthropic import ChatAnthropic
from langgraph.prebuilt import create_react_agent

deployment = metorial.provider_deployments.create(
    name="Metorial Search",
    provider_id="metorial-search",
)

session = await metorial.connect(
    adapter=metorial_langchain(),
    providers=[{"provider_deployment_id": deployment.id}],
)
llm = ChatAnthropic(model="claude-sonnet-4-20250514")
agent = create_react_agent(llm, session.tools())

result = await agent.ainvoke(
    {"messages": [("user", "Search the web for the latest news about AI agents and summarize the top 3 stories.")]}
)
print(result["messages"][-1].content)

PydanticAI

from metorial import metorial_pydantic_ai
from pydantic_ai import Agent

deployment = metorial.provider_deployments.create(
    name="Metorial Search",
    provider_id="metorial-search",
)

session = await metorial.connect(
    adapter=metorial_pydantic_ai(),
    providers=[{"provider_deployment_id": deployment.id}],
)
agent = Agent("anthropic:claude-sonnet-4-20250514", tools=session.tools())

result = await agent.run("Search the web for the latest news about AI agents and summarize the top 3 stories.")
print(result.output)

Error Handling

from metorial import (
    Metorial,
    AuthenticationError,
    NotFoundError,
    RateLimitError,
    OAuthRequiredError,
    metorial_openai,
)

metorial = Metorial()

try:
    session = await metorial.connect(
        adapter=metorial_openai(),
        providers=[{"provider_deployment_id": "your-deployment-id"}],
    )
    tools = session.tools()
except AuthenticationError:
    print("Check your METORIAL_API_KEY")
except NotFoundError:
    print("Deployment not found - verify your deployment ID")
except OAuthRequiredError:
    print("This provider requires OAuth - see the OAuth section above")
except RateLimitError:
    print("Rate limited - try again later")

License

MIT License - see LICENSE for details.

Support

Documentation · GitHub Issues · Email Support