Strip thinking tags API

Strips chain-of-thought / reasoning blocks (<thinking>, <think>, <scratchpad>, <reflection>, <reasoning>) from a response and returns the clean answer plus the extracted reasoning separately, with counts of tags and characters removed. Answers 'how do I remove the model's thinking from the final answer?', 'how do I separate reasoning from the response?'.

Price$0.02per request

MethodPOST

Route/v1/llm/strip-thinking

StatusLive

MIME typeapplication/json

Rate limit120/minute

Cache0s public

llmthinkingreasoningchain-of-thoughtstripcleancotagent

API URLhttps://x402.hexl.dev/v1/llm/strip-thinking

Integration docs

Example request

{
  "text": "<thinking>let me reason</thinking>\nThe answer is 7."
}

Example response

{
  "clean": "The answer is 7.",
  "thinking": [
    "let me reason"
  ],
  "removedTags": 1,
  "removedChars": 35
}

Input schema

{
  "type": "object",
  "required": [
    "text"
  ],
  "properties": {
    "text": {
      "type": "string",
      "examples": [
        "<thinking>let me reason</thinking>\nThe answer is 7."
      ]
    },
    "tags": {
      "type": "array",
      "items": {
        "type": "string"
      },
      "default": [
        "thinking",
        "think",
        "scratchpad",
        "reflection",
        "reasoning"
      ]
    }
  }
}

Output schema

{
  "type": "object",
  "additionalProperties": true
}