Runstack is the tool layer for AI agents, giving AI applications access to thousands of real-world tools and integrations without overwhelming the model with unnecessary context.

Building capable AI agents requires more than just an LLM. Agents need access to the applications, data, and services that businesses already use. But integrating every service individually means dealing with APIs, authentication, permissions, tool schemas, maintenance, and increasingly large context requirements.

Runstack provides a unified platform for discovering, connecting, and executing tools, allowing agents to work across a large ecosystem of applications while keeping their context focused on the task they are solving.

Thousands of Tools & Integrations

Runstack provides a growing ecosystem of toolkits across engineering, productivity, communication, cloud infrastructure, CRM, analytics, support, business operations, and more.

The ecosystem is continuously expanding, with support for new tools and integrations added through Runstack and its open-source toolkit ecosystem. Users can also request integrations or additional capabilities.

Intelligent Tool Discovery

Instead of requiring an AI model to carry every available tool definition in its context, Runstack enables agents to discover tools on demand.

An agent can find the capability it needs, understand how to use it, connect the required account, and perform the action. This allows agents to work with a much larger tool ecosystem while keeping their active context focused.

The result is a more scalable approach to agent tooling that can help reduce unnecessary input tokens, improve tool selection, and make agents faster and easier to build.

Runstack Agent

Runstack Agent provides a complete AI workspace for interacting with connected applications through natural language.

Users can chat with AI agents, connect their accounts, upload documents, use web search, interact with external MCP servers, and execute actions across their connected tools.

Runstack supports multiple AI model providers and allows teams to bring their own model credentials. The workspace also provides visibility into tool calls, results, reasoning, and token usage, making it easier to understand what an agent is doing.

Runstack Studio

Runstack Studio is a visual environment for building reusable AI agents and workflows.

Teams can define an agent's role, instructions, tools, workflow steps, and conditions, then publish and reuse those agents across their organization.

Studio supports reusable workflows for areas such as engineering, sales, support, operations, finance, and business automation. Published agents can be discovered through the marketplace and invoked directly from chat.

MCP Support

Runstack provides native MCP support, allowing its tool ecosystem to be used across modern AI development environments.

Developers can connect Runstack with Cursor, VS Code, Claude Desktop, Claude Code, ChatGPT, Codex CLI, Antigravity, and other MCP-compatible clients.

This gives developers a single way to bring a broad range of integrations into the AI tools and development environments they already use.

Developer SDKs

Runstack provides multiple SDKs and integration options depending on how developers want to use it.

@rnsk/tools brings Runstack's tools directly into applications using the Vercel AI SDK, supporting text generation, streaming, agent loops, and combinations with custom application tools. Developers can use the full ecosystem or scope agents to specific toolkits and workflows.

@rnsk/mcp provides a simple way to connect Runstack with MCP-compatible clients and development environments.

Authentication & Security

Runstack simplifies authentication across integrations by handling account connections and OAuth flows for supported applications.

Credentials remain outside the AI model while Runstack manages connections and tool execution. This allows agents to perform real actions without exposing sensitive authentication information to the model.

Organizations & Workspaces

Runstack supports both individual users and teams through organizations and workspaces.

Teams can invite members, manage multiple workspaces, share agents and connections, configure API keys and model providers, define workspace instructions, and manage security and access controls.

Each workspace can also have its own branding, making Runstack suitable for internal AI platforms as well as customer-facing AI products.

Open & Extensible Ecosystem

Runstack's toolkit ecosystem is designed to grow with the AI developer community.

Developers can contribute new integrations, extend existing toolkits, create specialized capabilities, and build agents around their own workflows.

Runstack — give your AI agents the tools to discover, connect, and act.

• Hosted server: runstack.engineer

• MCP endpoint: app.runstack.engineer/mcp clients use POST

• Documentation: runstack.engineer/docs

• OSS toolkit registry github: deepraj21/rnsk-toolkits & package at @rnsk/toolkits