Skip to main content
Runlayer connects to your agents two ways — pick the one that fits, or use both.

Connect your agent to MCP servers

Point an agent framework at a Runlayer MCP proxy URL. Your agent calls MCP tools as usual; Runlayer applies policy, scanning, and identity in the middle.

Claude Agent SDK

Native MCP support via query() and allowedTools, in TypeScript and Python.

Vercel AI SDK

MCP tools for generateText and streamText over HTTP or stdio.

OpenAI Agents SDK

Streamable HTTP, stdio, and hosted MCP tools for Python agents.

Google ADK

MCP servers in ADK agents, plus the MCP Toolbox for 40+ databases.

Govern and monitor what agents do

Stream an agent’s full activity — prompts, reasoning, tool calls, and responses — into Runlayer Sessions, with scanning, policy, and AgentGuard applied across the trajectory.

Hooks SDK

Instrument custom TypeScript and Python agents — stream the whole conversation into Sessions and enforce tool calls, with adapters for Claude, Vercel, OpenAI, and Google harnesses.

Claude Cowork monitoring

Export organization-wide Claude Cowork OpenTelemetry session logs to Runlayer — no per-machine hook or daemon required.