Validate LLM output: repair JSON, enforce schema, check grounding against a source, extract citations
POST /v1/agent-output-validator on agent.caelaralabs.com
Request parameters
| Name | In | Type | Required | Description |
|---|---|---|---|---|
output | body | string | yes | Raw LLM output |
schema | body | object | no | JSON Schema subset to enforce |
source | body | string | no | Source text for grounding check |
Call it with an API key
Send your key as X-Caelara-Key. Responses are JSON; when an upstream source is unavailable the API still answers HTTP 200 with "partial": true instead of a server error.
Use it as an MCP tool
Tool name agent_output_validator on the remote server https://agent.caelaralabs.com/mcp (registry ID com.caelaralabs/agent). MCP tool calls need an X-Caelara-Key API key; to pay per call in USDC, use the direct HTTP route in the x402 section below.
Pay per call in USDC on Base
No account needed. An unauthenticated request returns HTTP 402 with an X-Payment-Required: x402 header and a JSON body whose accepts entry carries scheme exact, network base, maxAmountRequired 25000 (0.025 USDC, 6 decimals), the USDC asset and the payTo address. Settlement happens only after a successful response.