AI agents for your product, hosted and ready to connect.

Create an agent, give it your documents and choose a model. Then connect it to your app, website or backend over REST or WebSocket.

Book a demo Log in

Manage agents from the dashboard, Claude, or anything that speaks MCP or REST.

How it works

Three steps from idea to a live agent.

No servers, no vector database and no infrastructure to run.

Create an agent

Give it a name, instructions and a model. Each agent can use its own model, chosen from the foundation models available on AWS Bedrock.

Add what it should know

Upload your documents and the agent answers from them. There's nothing to index, chunk or tune.

Connect it

Call it over REST, stream over WebSocket or drop the chat widget onto your site. Every conversation is saved so you can review it later.

Features

Everything you need to run agents in production.

Formerly Workflow Caddie, now focused on one job: hosting the agents your product talks to. Scheduled jobs and code execution live in Code Caddie.

📄

Knowledge bases without the plumbing

Organise documents into folders, and share a time-limited upload link with anyone who needs to add files. Agent Caddie decides how much of the library each answer needs.

⇄

Routers for teams of agents

Put several specialist agents behind one address. A router reads each message and passes it to the agent best suited to answer.

🔒

Keys you can hand out safely

Give an integrator a one-time claim link instead of pasting a secret. Keys can be limited to specific agents and revoked at any time.

❖

Manage it from Claude

Create agents, edit their knowledge and talk to them from Claude through the Agent Caddie connector, or from the dashboard.

⚙

Your code stays in control

Agents don't reach out to other systems on their own. They return a clear decision, and your code takes the action.

☰

Every conversation on record

Conversations are stored per agent, so you can review what was asked, what was answered and where things went wrong.

For developers

One request to talk to an agent.

Every agent has a REST endpoint and a WebSocket address. Send a message and get the agent's answer back, with a conversation ID that carries context across turns.

REST for request and response
WebSocket for streamed replies
Chat widget with one script tag
MCP connector at mcp.agentcaddie.com

# Send a message over REST curl -X POST https://api.agentcaddie.com/chat/$AGENT/$CHAT \ -H "Authorization: Bearer $AGENT_CADDIE_KEY" \ -H "Content-Type: application/json" \ -d '{"message": "How do I reset my password?"}' # Reply { "reply": "Go to Settings, then…", "conversation_id": "5131c5f1…" } # Or stream it over WebSocket wss://$AGENT.agentcaddie.com { "action": "sendmessage", "message": "…" }