CrypTok AI/SI API

Build with CrypTok's in-house AI/SI engine. OpenAI-compatible response shape, one model (Lumen) running on the team's own GPU server (rgb).

Base URL: https://cryptok.ai/api/v1Auth: Bearer API keyFormat: JSON or streaming text/plain

Quick start

Three steps to your first response:

API keys come with AI/SI Pro or a Killer Whale membership on cryptok.me.
Generate an API key at /upgrade — copy it once (shown only at creation).
Hit the endpoint:

curl

curl https://cryptok.ai/api/v1/chat \
  -H "Authorization: Bearer cak_<keyId>.<secret>" \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [
      {"role": "user", "content": "What is $CRYPTOK?"}
    ],
    "model": "cryptok-lumen"
  }'

Response is an OpenAI-style chat.completion object — drop it into existing tooling that expects that shape.

Authentication

Every request requires Authorization: Bearer cak_<keyId>.<secret> where the key was generated for an account with API access (AI/SI Pro or a Killer Whale membership). The secret half is hashed at rest (SHA-256), only the prefix is recoverable — store the full key in your env vars at creation time.

Key format

text

cak_<keyId16>.<secret32>

Example:
cak_a17c07022795d3a3.177a41a921924696175292845bbbdf35
└─┬─┘└──────┬───────┘└──────────────┬─────────────────┘
  prefix   public id (16 hex)        secret (32 hex)

Limits

Maximum 10 active keys per account — revoke unused keys at /upgrade
Keys are per-user not per-app — each call counts against your account's daily message limit
Revoked keys return 401 immediately (no grace period)

POST /api/v1/chat

Send messages, get a response. Stateless — you manage conversation history client-side and pass the full messages array each request.

Request body

json

{
  "messages": [
    {"role": "system", "content": "(optional) Custom system prompt"},
    {"role": "user", "content": "What is $CRYPTOK?"},
    {"role": "assistant", "content": "(optional) Prior assistant turn"},
    {"role": "user", "content": "Tell me more about its tokenomics."}
  ],
  "model": "cryptok-lumen",
  "stream": false
}

Field	Type	Required	Description
`messages`	array	yes	Conversation turns. Must include at least one `role: "user"`. Roles: `user`, `assistant`, `system`.
`model`	string	no	One of the 7 model ids below. Defaults to `cryptok-lumen`.
`stream`	boolean	no	If `true`, response is `text/plain` streaming of tokens as they generate. Default `false` = JSON.

Response (JSON mode)

json

{
  "id": "chatcmpl-s46v1rjzx6g",
  "object": "chat.completion",
  "created": 1780525913,
  "model": "cryptok-lumen",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "$CRYPTOK is the native Solana token..."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_chars": 17,
    "completion_chars": 412,
    "credits_charged": 3,
    "source": "quota"
  },
  "knowledge": {
    "hits": 5,
    "top_score": 0.81
  }
}

The knowledge object reports how many sanitized snippets from CrypTok's internal knowledge base were retrieved and the top similarity score. usage.source is either "quota" (drawn from your daily allowance) or "credits" (drawn from your credit balance).

Models

There is one model: cryptok-lumen (Lumen), the same engine every CrypTok AI/SI user chats with. The model field is optional; older ids (cryptok-lumen-pro, -nova, -forge, -spark, -coder-mini, -comet, -pico) are still accepted and run on Lumen.

Model id	Context	Best for
`cryptok-lumen`	32K tokens	Everything — conversation, research, code, crypto and on-chain questions.

Streaming

Pass "stream": true in the request to receive tokens as a text/plain stream — same format as the cryptok.ai web UI uses. Time to first byte is typically ~200ms even on Lumen.

javascript

const res = await fetch('https://cryptok.ai/api/v1/chat', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer cak_<keyId>.<secret>',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    messages: [{role: 'user', content: 'Explain Solana validators briefly.'}],
    model: 'cryptok-lumen',
    stream: true,
  }),
});

const reader = res.body.getReader();
const decoder = new TextDecoder();
while (true) {
  const { value, done } = await reader.read();
  if (done) break;
  process.stdout.write(decoder.decode(value, {stream: true}));
}

Response headers (live quota tracking)

Every successful response includes headers so you can show live usage in your app without a separate quota call:

Header	Type	Example
`X-CrypTok-AI-Used`	int	`1247` messages used today
`X-CrypTok-AI-Cap`	int	`5000` daily cap
`X-CrypTok-AI-Credits`	float	`247.5` credits in balance
`X-CrypTok-AI-Cost`	float	`3` credits this call cost
`X-CrypTok-AI-Source`	enum	`quota` or `credits`
`X-Request-Id`	string	`chatcmpl-s46v1rjzx6g`

Errors

Errors are returned as JSON with an error object containing a code and message:

json

{
  "error": {
    "code": "rate_limited",
    "message": "Daily cap reached",
    "used": 5000,
    "cap": 5000,
    "credit_balance": 0
  }
}

Code	HTTP	When	How to fix
`invalid_api_key`	401	Authorization header missing or key is malformed/revoked.	Send Authorization: Bearer cak_<keyId>.<secret>. Verify the key on /upgrade.
`tier_required`	403	Your account doesn't include API access.	API keys come with AI/SI Pro or a Killer Whale membership (cryptok.me/packages).
`bad_request`	400	Invalid JSON in request body.	Ensure Content-Type: application/json and the body is valid JSON.
`no_messages`	400	messages[] is empty or has no user-role message.	Include at least one {"role": "user", "content": "..."} entry.
`rate_limited`	429	Daily message limit reached.	Wait for the rolling 24h refill, or move to AI/SI Pro (250 messages a day). The error body includes used/cap.
`engine_error`	502	Inference backend (rgb) returned an error or timed out.	Retry with backoff. Persistent failure → contact support.

Rate limits

One model, one plan. Every call counts as one message against the key owner's daily limit — the same limits as the cryptok.ai chat.

Account	Messages / day	Images / day	API keys
Free CrypTok account	25	3	No
AI/SI Pro ($29/mo · $78/3 mo · $290/yr)	250	30	✓ Yes
Degen membership (includes AI/SI Pro limits)	250	30	No
Killer Whale membership (includes AI/SI Pro)	250	30	✓ Yes

Limits refill on a rolling 24-hour window, not calendar-day midnight. When they run out you get a 429 rate_limited error until they refill.

SDK examples

Node.js / TypeScript

typescript

async function chat(message: string): Promise<string> {
  const res = await fetch('https://cryptok.ai/api/v1/chat', {
    method: 'POST',
    headers: {
      'Authorization': `Bearer ${process.env.CRYPTOK_AI_KEY}`,
      'Content-Type': 'application/json',
    },
    body: JSON.stringify({
      messages: [{role: 'user', content: message}],
      model: 'cryptok-lumen',
    }),
  });
  if (!res.ok) {
    const err = await res.json();
    throw new Error(`${err.error.code}: ${err.error.message}`);
  }
  const data = await res.json();
  return data.choices[0].message.content;
}

console.log(await chat('Whats the price of SOL?'));

Python

python

import os
import requests

def chat(message: str) -> str:
    res = requests.post(
        'https://cryptok.ai/api/v1/chat',
        headers={
            'Authorization': f'Bearer {os.environ["CRYPTOK_AI_KEY"]}',
            'Content-Type': 'application/json',
        },
        json={
            'messages': [{'role': 'user', 'content': message}],
            'model': 'cryptok-lumen',
        },
        timeout=120,
    )
    res.raise_for_status()
    data = res.json()
    return data['choices'][0]['message']['content']

print(chat('Whats the price of SOL?'))

OpenAI SDK (drop-in replacement)

The response shape is compatible with the OpenAI chat.completion object, so you can use the official OpenAI SDK by changing the baseURL:

typescript

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: process.env.CRYPTOK_AI_KEY,
  baseURL: 'https://cryptok.ai/api/v1',
});

const completion = await client.chat.completions.create({
  model: 'cryptok-lumen',
  messages: [{role: 'user', content: 'Hello!'}],
});

console.log(completion.choices[0].message.content);

Note: OpenAI SDK quirks like tools array, function calls, response_format: json, and seed may behave differently or not be supported. Stick with the simple messages → choices path for guaranteed compatibility.