How to Use the Weights & Biases MCP in AutoGen

Q: Can AutoGen help decide between projects using Weights & Biases?

Yes. Agents can use listwandbprojects to gather all available projects. They then negotiate which project best fits a complex criteria set, acting like an internal review board.

Q: What is the data privacy scope when using Weights & Biases and AutoGen?

The server exposes metadata types, such as artifact names (listprojectartifacts) and run IDs. The agents only debate these visible identifiers, maintaining separation from private compute outputs.

Q: How do I check historical reports using Weights & Biases and AutoGen?

Agents can use listprojectreports to pull the names of all saved analysis reports. They then debate which report is most relevant based on a user's complex query.

Q: Does Weights & Biases provide structure for AutoGen?

The server provides structured tools like listprojectsweeps. The agents use the list output to initiate debates, ensuring that the decision-making process is grounded in actual W&B experiment data.

Drive complex decisions with AutoGen and weights-biases-mcp multi-agent debate.

See Vinkius in Action

Works with every AI agent you already use

…and any MCP-compatible client

MCP Servers - Free for Subscribers

Connect Weights & Biases MCP to AutoGen

Create your Vinkius account to connect Weights & Biases to AutoGen and route execution through our secure gateway. The platform manages server hosting, runtime updates, and security layers. Configuration requires no manual server provisioning.

GDPR Free for Subscribers

Setup Weights & Biases with AutoGen

Ask AI about this MCP

ChatGPT

Claude

Perplexity

Debating Experiment Metrics

The `get_run_details` tool provides the foundational data for an argument. Multiple agents can use this information—the 'performance agent' checks metrics, while a 'risk agent' reviews the config. They debate which run is optimal based on different criteria, converging on a single decision point.

Collaborative Resource Discovery

Agents can use `list_project_artifacts` to discover available resources. A 'data agent' might find a dataset while an 'architecture agent' finds the corresponding model. This list becomes part of their negotiation. This multi-step process mimics human collaboration, forcing consensus on necessary components.

Challenging Hyperparameter Choices

When `list_project_sweeps` runs, the agent system gets a list of all attempts. The agents then debate which sweep pattern was best, and why. They challenge assumptions based on the sheer volume of data presented. This is ideal for building decision systems where the answer requires multiple expert perspectives.

Setup guide

Set up Weights & Biases MCP in AutoGen

Prerequisites

Python 3.10+ installed
autogen-ext[mcp] package
Active Vinkius subscription with a valid endpoint token

1

Install AutoGen with MCP
Run pip install "autogen-ext[mcp]" autogen-agentchat. The MCP extension includes mcp_server_tools for stateless tool access.
2

Fetch tools from the MCP
Call mcp_server_tools(SseServerParams(url=...)) with your Vinkius endpoint. Replace [YOUR_TOKEN_HERE] with your token from cloud.vinkius.com.
3

Run your agent
Pass the tools to AssistantAgent and call agent.run(). The agent invokes Weights & Biases tools and returns structured results.

agent.py

from autogen_ext.tools.mcp import SseServerParams, mcp_server_tools
from autogen_agentchat.agents import AssistantAgent
from autogen_ext.models.openai import OpenAIChatCompletionClient

server_params = SseServerParams(
    url="https://edge.vinkius.com/[YOUR_TOKEN_HERE]/mcp"
)

tools = await mcp_server_tools(server_params)

agent = AssistantAgent(
    name="Weights & Biases_assistant",
    model_client=OpenAIChatCompletionClient(model="gpt-4o"),
    tools=tools,
)

result = await agent.run("List recent Weights & Biases data")
print(result.messages[-1].content)

Get your connection token →

Prerequisites

Python 3.10+ installed
autogen-ext[mcp] + autogen-agentchat
Active Vinkius subscription with a valid endpoint token

1

Install dependencies
Same packages as above. McpWorkbench is ideal when your agent needs stateful sessions across multiple tool calls.
2

Use McpWorkbench as context manager
Wrap your agent in async with McpWorkbench(...) to maintain shared state and resources. The workbench manages the full MCP session lifecycle.
3

Run with workbench
Pass workbench=workbench to your agent. State is preserved across multiple tool calls within the same session.

agent.py

from autogen_ext.tools.mcp import McpWorkbench, SseServerParams
from autogen_agentchat.agents import AssistantAgent
from autogen_ext.models.openai import OpenAIChatCompletionClient

server_params = SseServerParams(
    url="https://edge.vinkius.com/[YOUR_TOKEN_HERE]/mcp"
)

async with McpWorkbench(server_params) as workbench:
    agent = AssistantAgent(
        name="Weights & Biases_assistant",
        model_client=OpenAIChatCompletionClient(model="gpt-4o"),
        workbench=workbench,
    )

    result = await agent.run("List recent Weights & Biases data")
    print(result.messages[-1].content)

Get your connection token →

Independent Platform Disclaimer: Vinkius is an independent platform and is not affiliated with, endorsed by, sponsored by, verified by, or otherwise authorized by Weights & Biases. All third-party trademarks, logos, and brand names are the property of their respective owners. Their use on this website is strictly for informational purposes to identify service compatibility and interoperability.

Why Choose Vinkius

Vinkius connects your tools to AI with real-time monitoring and automatic cost savings — all from one dashboard.

Connect Weights & Biases now

Real-time monitoring

Live

visibility into every interaction

Connect your favorite tools to your AI and see exactly what's happening — every request, every response, in real time.

Built-in savings

60%

lower AI costs

Vinkius compresses data between your apps and your AI automatically. Lower bills every month — no configuration required.

Single dashboard

One

place for every integration

Every tool your AI connects to, managed from a single screen. One account, complete control.

Common questions about Weights & Biases MCP in AutoGen

The MCP Server exposes W&B tools as inputs to the agents. Instead of one agent calling a tool, multiple specialized agents debate the results shown by the `list_project_runs` output.

Yes. Agents can use `list_wandb_projects` to gather all available projects. They then negotiate which project best fits a complex criteria set, acting like an internal review board.

The server exposes metadata types, such as artifact names (`list_project_artifacts`) and run IDs. The agents only debate these visible identifiers, maintaining separation from private compute outputs.

Agents can use `list_project_reports` to pull the names of all saved analysis reports. They then debate which report is most relevant based on a user's complex query.

The server provides structured tools like `list_project_sweeps`. The agents use the list output to initiate debates, ensuring that the decision-making process is grounded in actual W&B experiment data.

Use it with your favorite AI tools

Connect this server to Cursor, Claude, VS Code, and more.

OpenAI Agents SDK sdk-python

Google ADK sdk-python

Pydantic AI sdk-python

Vercel AI SDK sdk-typescript