Tenant-scoped OpenAI and Anthropic-compatible endpoints, chaos injection, and MCP support. No API key. No rate limits. Just HTTP.
Each tenant has its own mock state and scenario library. Pass X-Tenant-ID in every request.
Use /v1/chat/completions (OpenAI) or /v1/messages (Anthropic), or send MCP JSON-RPC to /mcp/message.
Responses stream token-by-token from a scenario matched by keyword. Inject chaos to simulate errors or latency.
Connect Claude Code, Cursor, or any MCP-compatible editor to your tenant. Your agent can create draft scenarios — you review and approve before they go live.
Get an API token from Playground → API Tokens.
Add the MCP server to your editor config.
Your agent calls create_scenario — drafts appear in the Scenarios tab for your review.
{
"mcpServers": {
"llmplaceholder": {
"type": "http",
"url": "https://llmplaceholder.com/mcp/message",
"headers": {
"X-Tenant-ID": "my-tenant",
"Authorization": "Bearer <api-token>"
}
}
}
}