Toolhouse
Last updated
Last updated
Toolhouse is an agentic infrastructure and development stack. Toolhouse allows you to define agents and deploy them into APIs. The platform has built-in eval, memory, RAG, code execution, browser use, and all the functionality you need to build agentic workflows.
With Toolhouse, you don't have to write boilerplate or buy and integrate separate parts of the stack, but you can still do so if you want to bring your own.
Toolhouse allows you to schedule agents to run at specific intervals; your agents can also trigger webhooks or call URLs you specify.
Toolhouse is an open platform that supports MCP, allowing you to keep your data, resources, and agents model-independent. Toolhouse is both an MCP client and a server. You can connect existing MCP servers and run them in the Toolhouse infrastructure; you can also build your MCP servers and even make them public.
Agent Studio allows you to build agents with simple instructions. Toolhouse optimizes the prompt for each agent and extends its functionality with built-in MCP servers. You can also upload and manage knowledge and data by simply uploading files to the agent.
You can deploy any Agent Studio agent as an API with a dedicated endpoint. When done, your agent can autonomously call a webhook, a callback, or another agent.
You can schedule agents to run a specific interval.
You can inspect each agent run. Agent Logs show you the entire execution lifecycle of each agent you run. For a deeper drill down, you can use Execution logs to check the MCP servers called by your agents.
The Toolhouse SDK allows you to run Toolhouse in your existing codebase. You can use the SDK if you want to use your own custom agents. The SDK integrates with Vercel AI and LlamaIndex.