Agent Chat

Send messages to agents through the Shroud LLM proxy. Conversations persist across sessions with full message history.

Sending a message

// SDK
const response = await client.chat.sendMessage(agentId, {
  message: "What API keys do I have stored?",
  conversation_id: conversationId, // optional, creates new if omitted
  model: "gpt-4o",
  provider: "openai"
});

SSE streaming

const stream = await client.chat.sendMessageStream(agentId, {
  message: "Analyze my vault secrets",
});

const reader = stream.body.getReader();
// Process SSE events...

Runtime chat

Chat directly with agents running in cloud runtimes via an OpenAI-compatible in-container bridge:

POST /v1/runtimes/{id}/chat

Step-up auth may be required first:

POST /v1/runtimes/{id}/chat/unlock

The dashboard shows a Chat tab alongside the Terminal on runtime detail pages.

Conversations

List and manage conversations:

const conversations = await client.chat.listConversations(agentId);
const detail = await client.chat.getConversation(agentId, conversationId);
await client.chat.deleteConversation(agentId, conversationId);

MCP tools

Tool	Description
send_chat_message	Send a message to an agent and get a response
list_chat_conversations	List agent chat conversations

API endpoints

Method	Path	Description
POST	/v1/agents/{id}/chat	Send message (SSE streaming)
POST	/v1/agents/{id}/chat/unlock	Step-up auth to unlock chat
GET	/v1/agents/{id}/chat/conversations	List conversations
GET	/v1/agents/{id}/chat/conversations/{cid}	Get conversation
DELETE	/v1/agents/{id}/chat/conversations/{cid}	Delete conversation
POST	/v1/runtimes/{id}/chat	Runtime chat (SSE)
POST	/v1/runtimes/{id}/chat/unlock	Runtime chat unlock