Weavz
Agent frameworks

Use the framework you want. Let Weavz own the action layer.

Your framework decides how the agent plans, reasons, branches, and hands off. Weavz gives it scoped app access, user-owned credentials, MCP servers, durable state, files, approvals, and audit receipts.

OpenAI Agents SDK
Provider SDK
LangGraph
Graph orchestration
LangChain
Agent framework
CrewAI
Multi-agent teams
Google ADK
Provider SDK
Microsoft Agent Framework
Enterprise framework
AutoGen / AG2
Multi-agent runtime
SK
Semantic Kernel
Enterprise orchestration
LlamaIndex Workflows
Data workflows
Mastra
TypeScript agents
Vercel AI SDK
TypeScript AI apps
Pydantic AI
Python agents
Compatibility surface

Weavz plugs in where frameworks expect tools.

Use MCP when your harness can consume a remote server. Use the TypeScript or Python SDK when you want framework-native tool definitions backed by Weavz action execution.

OpenAI Agents SDK

Provider SDK

Build agent orchestration in code and attach Weavz as a hosted MCP tool source.

Use when

You build agents in OpenAI code and need a governed tool source.

LangGraph

Graph orchestration

Use Weavz MCP endpoints inside stateful graph workflows and long-running agent flows.

Use when

Your agent workflow is a graph with durable branches and state.

LangChain

Agent framework

Load Weavz tools through MCP adapters or route SDK tool calls back through Weavz.

Use when

You already use LangChain tools, adapters, or chains.

CrewAI

Multi-agent teams

Give crews controlled business actions without embedding customer credentials in prompts.

Use when

You coordinate role-based crews that need app access.

Google ADK

Provider SDK

Connect ADK agents to Weavz over Streamable HTTP MCP or generated SDK tools.

Use when

You are building with Google ADK and want hosted business tools.

Microsoft Agent Framework

Enterprise framework

Use Weavz as the governed action layer behind Microsoft-stack agent apps.

Use when

Your agents live in a Microsoft-stack application or service.

AutoGen / AG2

Multi-agent runtime

Attach scoped Weavz workspaces to conversational multi-agent systems.

Use when

You need multi-agent conversations with scoped business actions.

SK

Semantic Kernel

Enterprise orchestration

Call Weavz from planner and plugin-style workflows where business actions need audit.

Use when

You use planner or plugin-style enterprise orchestration.

LlamaIndex Workflows

Data workflows

Pair document-heavy agents with governed actions, files, state, browser, and approvals.

Use when

Your agent starts with documents, indexes, or retrieval workflows.

Mastra

TypeScript agents

Use a TypeScript-first agent stack while Weavz handles app access and user auth.

Use when

You want a TypeScript agent app with hosted tool execution.

Vercel AI SDK

TypeScript AI apps

Convert Weavz action tools into AI SDK tool sets for app-owned agent loops.

Use when

Your product loop is built with Vercel AI SDK tools.

Pydantic AI

Python agents

Keep Python agents type-safe while Weavz validates and executes app actions.

Use when

You want typed Python agents with validated tool inputs.

Agno

Python agents

Run lightweight Python agents against Weavz-hosted app actions and state.

Use when

You are shipping compact Python agents and need external actions.

Haystack

RAG pipelines

Add governed actions to retrieval and document pipelines when agents need to act.

Use when

Your pipeline retrieves information and then needs to take action.

smolagents

Code agents

Give code-first agents a scoped MCP endpoint instead of raw customer credentials.

Use when

You want lightweight code agents with a narrow action boundary.

DSPy

Programmatic LM apps

Keep prompt/program optimization separate from the runtime that touches business apps.

Use when

You optimize language programs but do not want raw credentials in them.

Claude Agent SDK

Provider SDK

Use Weavz through MCP when Claude-based agents need customer app access.

Use when

You are building Claude-based agents and want remote MCP tools.

Custom MCP clients

Custom harnesses

Any Streamable HTTP MCP client can receive a workspace endpoint and scoped mcp_ token.

Use when

You built your own MCP client, IDE harness, or agent runtime.

Starter snippets

Hand the harness an endpoint, not a pile of app credentials.

Your backend creates the workspace, end user, connect links, MCP server, and scoped token first. These snippets show only the framework handoff.

Generic MCP harness
json
{
  "mcpServers": {
    "weavz-acme": {
      "type": "http",
      "url": "https://platform.weavz.io/mcp/srv_...",
      "headers": {
        "Authorization": "Bearer mcp_..."
      }
    }
  }
}
OpenAI Agents SDK
typescript
import { Agent, hostedMcpTool } from "@openai/agents";

const agent = new Agent({
  name: "Customer ops agent",
  tools: [
    hostedMcpTool({
      serverLabel: "weavz",
      serverUrl: "https://platform.weavz.io/mcp/srv_...",
      headers: { Authorization: "Bearer mcp_..." },
    }),
  ],
});
LangChain / LangGraph
python
from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent

client = MultiServerMCPClient({
    "weavz": {
        "transport": "http",
        "url": "https://platform.weavz.io/mcp/srv_...",
        "headers": {"Authorization": "Bearer mcp_..."},
    }
})

tools = await client.get_tools()
agent = create_react_agent(model, tools)
Google ADK
python
from google.adk.tools.mcp_tool.mcp_toolset import McpToolset
from google.adk.tools.mcp_tool.mcp_session_manager import StreamableHTTPConnectionParams

weavz_tools = McpToolset(
    connection_params=StreamableHTTPConnectionParams(
        url="https://platform.weavz.io/mcp/srv_...",
        headers={"Authorization": "Bearer mcp_..."},
    )
)

Provision first

Create workspaces, users, aliases, and connect links from your backend.

Use MCP for agents

Give the framework a remote MCP endpoint plus a scoped mcp_ token.

Keep control

Human Gates, state, files, triggers, and audit remain in the workspace.

Direct personal trials use the generic connector URL https://platform.weavz.io/mcp/weavz. Builder harnesses usually create workspace-specific MCP servers and per-user tokens through the API or SDK.