MCP Server

Active

Apache-2.0

ai.klavis/strata MCP Server

ai.klavis/strata

One MCP gateway that lets AI agents progressively use thousands of tools without blowing up the context window.

What is the ai.klavis/strata MCP server?

Strata (ai.klavis/strata) is an MCP server by Klavis AI that acts as an intelligent gateway/connector layer for AI agents, letting them progressively access thousands of tools across many MCP servers without overloading the context window. It aggregates multiple MCP servers (e.g. Gmail, Slack, GitHub, Playwright) behind a single endpoint, exposing tools on-demand rather than dumping the full tool list to the model.

Strata solves the problem of context-window bloat and tool-overload when an AI agent needs access to many tools at once. Instead of registering every tool from every integration up front, Strata provides "progressive tool usage" so agents can discover and invoke tools from multiple underlying MCP servers as needed, at scale. It can be used as a hosted remote MCP endpoint (https://strata.klavis.ai/mcp/), self-hosted via pipx/Docker, or provisioned programmatically via the Klavis Python/TypeScript SDKs and REST API, which let you spin up a Strata instance backed by a chosen set of servers (e.g. Gmail, Slack).

How to install ai.klavis/strata

Copy-paste configuration for popular MCP clients.

transport: http

Config generated by PluginBench — verify against the source before use.

Cursor

~/.cursor/mcp.json

{
  "mcpServers": {
    "strata": {
      "url": "https://strata.klavis.ai/mcp/"
    }
  }
}

Windsurf

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "strata": {
      "serverUrl": "https://strata.klavis.ai/mcp/"
    }
  }
}

VS Code

.vscode/mcp.json

{
  "servers": {
    "strata": {
      "type": "http",
      "url": "https://strata.klavis.ai/mcp/"
    }
  }
}

Claude Desktop

Claude Desktop connects to remote servers via the mcp-remote bridge.

{
  "mcpServers": {
    "strata": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://strata.klavis.ai/mcp/"
      ]
    }
  }
}

Claude Code

claude mcp add --transport http strata https://strata.klavis.ai/mcp/

Use cases

Combine multiple third-party integrations (e.g. Gmail, Slack, GitHub) behind a single MCP endpoint for an agent
Reduce context-window usage by progressively exposing tools instead of loading all tools upfront
Provision a Strata server programmatically via SDK/REST API tied to a specific user and set of backend servers
Self-host a local Strata instance and attach stdio-based MCP servers like Playwright

ai.klavis/strata MCP server FAQ

What does the Strata MCP server do?

It acts as an aggregating gateway that lets AI agents access tools from multiple underlying MCP servers (like Gmail, Slack, GitHub, Playwright) through one connection, exposing tools progressively to avoid overloading the agent's context window.

Is Strata free to use?

Strata is open source and can be self-hosted for free via pipx or Docker. Klavis also offers a cloud-hosted version and API/SDK access which may require an API key and could involve paid usage tiers.

How do I connect Strata in Cursor or Claude?

Use the remote streamable-http endpoint https://strata.klavis.ai/mcp/ as the MCP server URL in your client's MCP configuration, or self-host it locally with 'pipx install strata-mcp' and 'strata add' to attach stdio servers.

Does Strata require authentication?

The cloud-hosted and SDK/REST API options require a Klavis API key. The remote streamable-http endpoint and self-hosted local install do not necessarily require an API key for basic use, depending on configuration.

Can Strata combine multiple MCP servers into one?

Yes, via the Klavis SDK or REST API you can create a Strata server instance specifying a list of servers (e.g. GMAIL, SLACK) that get bundled behind one Strata MCP endpoint for a given user.

What is the difference between Strata and Klavis's other offerings?

Strata is the intelligent tool-aggregation connector layer, while Klavis also offers 100+ prebuilt individual MCP integrations with OAuth support and an MCP Sandbox for scalable LLM training/RL environments.

README (reference)

Source of truth, from the repository.

</div>

🎯 Choose Your Solution

<div align="center"> <table> <tr> <td align="center" width="33%" valign="top" style="vertical-align: top; height: 250px;"> <div style="height: 100%; display: flex; flex-direction: column; justify-content: space-between;"> <div> <h2>Strata</h2> <p><strong>Intelligent connectors for your AI agent, optimize context window</strong></p> </div> <div> <a href="https://www.klavis.ai/docs/concepts/strata"> <img src="https://img.shields.io/badge/Explore-Strata-blue?style=for-the-badge&logo=data:image/svg+xml;base64,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" height="40"> </a> </div> </div> </td> <td align="center" width="33%" valign="top" style="vertical-align: top; height: 250px;"> <div style="height: 100%; display: flex; flex-direction: column; justify-content: space-between;"> <div> <h2>MCP Integrations</h2> <p><strong>100+ prebuilt integrations out-of-the-box, with OAuth support</strong></p> </div> <div> <a href="https://www.klavis.ai/docs/mcp-server/overview"> <img src="https://img.shields.io/badge/Explore-MCP%20Servers-purple?style=for-the-badge&logo=data:image/svg+xml;base64,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" height="40"> </a> </div> </div> </td> <td align="center" width="33%" valign="top" style="vertical-align: top; height: 250px;"> <div style="height: 100%; display: flex; flex-direction: column; justify-content: space-between;"> <div> <h2>MCP Sandbox</h2> <p><strong>scalable MCP environments for LLM training and RL</strong></p> </div> <div> <a href="https://www.klavis.ai/docs/concepts/sandbox"> <img src="https://img.shields.io/badge/Explore-Sandbox-orange?style=for-the-badge&logo=data:image/svg+xml;base64,PHN2ZyB3aWR0aD0iMjQiIGhlaWdodD0iMjQiIHZpZXdCb3g9IjAgMCAyNCAyNCIgZmlsbD0ibm9uZSIgeG1sbnM9Imh0dHA6Ly93d3cudzMub3JnLzIwMDAvc3ZnIj4KPHBhdGggZD0iTTQgNEgyMFY4TDE4IDEwTDIwIDE2VjIwSDRWMTZMNiAxMEw0IDhWNFoiIHN0cm9rZT0id2hpdGUiIHN0cm9rZS13aWR0aD0iMiIgc3Ryb2tlLWxpbmVjYXA9InJvdW5kIiBzdHJva2UtbGluZWpvaW49InJvdW5kIi8+CjxwYXRoIGQ9Ik00IDhIMjAiIHN0cm9rZT0id2hpdGUiIHN0cm9rZS13aWR0aD0iMiIvPgo8cGF0aCBkPSJNNCAxNkgyMCIgc3Ryb2tlPSJ3aGl0ZSIgc3Ryb2tlLXdpZHRoPSIyIi8+Cjwvc3ZnPg==" height="40"> </a> </div> </div> </td> </tr> </table> </div>

Quick Start

Option 1: Cloud-hosted - klavis.ai

Quickstart guide →

Option 2: Self-host

# Run any MCP Integration
docker pull ghcr.io/klavis-ai/github-mcp-server:latest
docker run -p 5000:5000 ghcr.io/klavis-ai/github-mcp-server:latest

# Install Open Source Strata locally
pipx install strata-mcp
strata add --type stdio playwright npx @playwright/mcp@latest

Option 3: SDK

# Python SDK
from klavis import Klavis
from klavis.types import McpServerName

klavis = Klavis(api_key="your-key")

# Create Strata instance
strata = klavis_client.mcp_server.create_strata_server(
    user_id="user123",
    servers=[McpServerName.GMAIL, McpServerName.SLACK],
)

# Or use individual MCP servers
gmail = klavis.mcp_server.create_server_instance(
    server_name=McpServerName.GMAIL,
    user_id="user123",
)

// TypeScript SDK
import { KlavisClient, McpServerName } from 'klavis';

const klavis = new KlavisClient({ apiKey: 'your-api-key' });

// Create Strata instance
const strata = await klavis.mcpServer.createStrataServer({
    userId: "user123",
    servers: [Klavis.McpServerName.Gmail, Klavis.McpServerName.Slack],
});

// Or use individual MCP servers
const gmail = await klavis.mcpServer.createServerInstance({
    serverName: McpServerName.GMAIL,
    userId: "user123"
});

Option 4: REST API

# Create Strata server
curl -X POST "https://api.klavis.ai/v1/mcp-server/strata" \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "user_id": "user123",
    "servers": ["GMAIL", "SLACK"]
  }'

# Create individual MCP server
curl -X POST "https://api.klavis.ai/v1/mcp-server/instance" \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "server_name": "GMAIL",
    "user_id": "user123"
  }'

Resources

Apache-2.0

View repository →