Skip to content
Weavz
Menu

Agent Framework and MCP Examples

Choose a native framework, AI client, or task workflow to connect agents to Weavz through MCP.

Use the agent framework you already have. Weavz gives it governed access to the integrations configured in a workspace, while your framework or AI client continues to handle the model.

Choose a starting point

You want to…Start here
Use Weavz from an existing AI clientConnect the hosted Weavz MCP App with OAuth, then choose the workspace the client should use.
Build an agent with a frameworkUse a framework starter with its native MCP client and a custom Weavz MCP server.
Start from a concrete taskAdapt a workflow example for inbox briefings, design reviews, sales reports, research, or another bounded task.

The hosted connector and the developer endpoint serve different setup paths. The connector at https://platform.weavz.io/mcp/weavz uses client OAuth. Framework starters connect to a custom MCP server and use an end-user scoped MCP access token issued by a trusted backend. Keep the customer REST API key on that backend. Do not place it in a model prompt or MCP client configuration.

Choose the MCP mode

Tool Mode exposes configured integration actions as separate MCP tools. It suits agents that need a small, explicit tool surface. Code Mode exposes discovery, API reading, and execution tools so an agent can inspect available actions and compose related calls. The server's mode determines which surface a client receives.

The hosted personal connector uses Code Mode. Custom framework servers can use Tool Mode or Code Mode. The examples below use each framework or client's native MCP support. Configure the workspace integrations, enabled actions, connections, and end-user identity that the task needs. The server exposes the access configured for that workspace and identity.

Choose by app and framework

Open the MCP integration and framework directory to find an app, then choose a framework setup recipe or a tested task workflow. The framework starters include 72 app setup recipes across 18 Python and TypeScript projects, with required actions, task inputs, and bounded example prompts.

App and frameworkSetup recipe
Slack MCP with LangChain PythonRead a channel through Weavz MCP
Gmail MCP with Google ADK PythonPrepare an inbox brief through Weavz MCP
GitHub MCP with Mastra TypeScriptInspect issues through Weavz MCP
Figma MCP with PydanticAI PythonReview design context through Weavz MCP

These recipes use the app integrations available on your Weavz MCP server. Connect your own app accounts before running a live task. For agents selecting a project automatically, examples-index.json records each example's apps, framework, language, authentication, MCP mode, and verification scope.

Framework starters

These projects show native MCP integration from common Python and TypeScript agent frameworks. Open a repository for its dependencies, setup steps, and local fixture checks.

Python

TypeScript

AI client setup

Use these projects when you want a client-specific MCP config and instruction files. The Weavz connector uses OAuth; the client stores its authorization. Model-provider setup stays in the AI client.

See the Weavz MCP App guide for OAuth, workspace selection, and the hosted endpoint. Each client repository explains its own project or profile scope. Follow the client’s current instructions for trusting a project and authorizing its MCP connection.

Task workflows

These PydanticAI projects provide scoped task instructions, an MCP connection, and a synthetic local fixture. They are starting points to adapt, not claims that a connected account has been reviewed or changed.

Browse the MCP examples hub for the full collection and contribution guidance. For a custom framework integration using the REST API or SDKs instead of MCP, start with the API and SDK quick start.