MoltSearch API

The MoltSearch API gives you access to 67 models from 9 providers through a single, unified interface — and, uniquely, to the debate engine that powers MoltSearch itself: ask once, have several models answer independently, and get one adjudicated verdict back.

The API surface

These are the endpoints we support today and intend to keep supporting. Anything not listed here is not part of the public API. Start with POST /v2/debate — the v1 endpoints below still work and are flat-priced, but they will be retired later, on notice. GET /v2/models and GET /v2/openapi.json need no key.

Quick Start

curl -X POST https://api.moltsearch.ai/v1/chat \
  -H "Content-Type: application/json" \
  -H "x-api-key: ms_live_YOUR_KEY_HERE" \
  -d '{
    "model": "claude",
    "message": "What is quantum computing?"
  }'

import requests

response = requests.post(
    "https://api.moltsearch.ai/v1/chat",
    headers={
        "Content-Type": "application/json",
        "x-api-key": "ms_live_YOUR_KEY_HERE"
    },
    json={
        "model": "claude",
        "message": "What is quantum computing?"
    }
)

data = response.json()
print(data["data"]["message"])

const response = await fetch("https://api.moltsearch.ai/v1/chat", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "x-api-key": "ms_live_YOUR_KEY_HERE"
  },
  body: JSON.stringify({
    model: "claude",
    message: "What is quantum computing?"
  })
});

const data = await response.json();
console.log(data.data.message);

payload := map[string]string{
    "model":   "claude",
    "message": "What is quantum computing?",
}
body, _ := json.Marshal(payload)

req, _ := http.NewRequest("POST", "https://api.moltsearch.ai/v1/chat", bytes.NewBuffer(body))
req.Header.Set("Content-Type", "application/json")
req.Header.Set("x-api-key", "ms_live_YOUR_KEY_HERE")

resp, _ := http.DefaultClient.Do(req)
defer resp.Body.Close()

Authentication

All API requests (except /v1/models and /v1/health) require an API key. Include your key in the x-api-key header:

x-api-key: ms_live_YOUR_KEY_HERE

Base URLs

Rate Limits

The default rate limit is 60 requests per minute per API key. If you exceed this, you'll receive a 429 response.

The Debate API (v2)

POST /v2/debate is the MoltSearch product. It is not a router and not a fan-out: it adjudicates. One question goes to several models independently, then a chairman model reads every answer and returns a single synthesised verdict — along with a ranking of the debaters, an agreement score that is measured rather than self-reported, and cited web sources.

Concretely, one call gives you:

• Independent answers. Each model answers without seeing the others, so you get genuine spread rather than an echo.
• A chairman verdict. One model reads every answer and writes the synthesis, citing sources when grounding is on.
• A ranking. The chairman orders the debaters best-first, in the same call as the verdict.
• An agreement score. Mean pairwise cosine similarity between embeddings of the answers. A number we computed, not one a model asserted. High agreement means the models converged; low agreement is your signal that the question is genuinely contested.
• Grounding. With grounded: true the panel is given live web results and the verdict carries [n] citations.

Run a debate

Requires an API key. Billed on the tokens actually used — see Billing.

Request body

Example request

curl -X POST https://api.moltsearch.ai/v2/debate \
  -H "Content-Type: application/json" \
  -H "x-api-key: ms_live_YOUR_KEY" \
  -d '{
    "query": "Is nuclear power essential to decarbonising the grid?",
    "models": [
      "openai/gpt-5-nano",
      "anthropic/claude-haiku-4-5",
      "google/gemini-2.5-flash-lite"
    ],
    "chairman": "anthropic/claude-sonnet-5",
    "grounded": true,
    "max_spend_credits": 2
  }'

Response

{
  "success": true,
  "data": {
    "query": "Is nuclear power essential to decarbonising the grid?",
    "verdict": {
      "text": "The models converge on nuclear being valuable but not strictly essential [1][3]...",
      "chairman": "anthropic/claude-sonnet-5",
      "chairman_model": "claude-sonnet-5"
    },
    "answers": [
      { "model": "openai/gpt-5-nano", "provider_model": "gpt-5-nano", "answer": "..." },
      { "model": "anthropic/claude-haiku-4-5", "provider_model": "claude-haiku-4-5-20251001", "answer": "..." },
      { "model": "google/gemini-2.5-flash-lite", "provider_model": "gemini-2.5-flash-lite", "answer": "..." }
    ],
    "rankings": [
      { "model": "anthropic/claude-haiku-4-5", "avgRank": 1 },
      { "model": "openai/gpt-5-nano", "avgRank": 2 },
      { "model": "google/gemini-2.5-flash-lite", "avgRank": 3 }
    ],
    "agreement_score": 0.8353,
    "sources": [ { "n": 1, "title": "...", "url": "https://..." } ],
    "models_unavailable": [],
    "models_substituted": [],
    "grounded": true
  },
  "usage":  { "credits_used": 0.3133, "credits_remaining": 41.33 },
  "meta": {
    "debaters": ["openai/gpt-5-nano", "anthropic/claude-haiku-4-5", "google/gemini-2.5-flash-lite"],
    "chairman": "anthropic/claude-sonnet-5",
    "billing": {
      "credits_charged": 0.3133,
      "credits_remaining": 41.33,
      "capped_by_max_spend": false,
      "usage_reported": true,
      "unpriced_models": []
    }
  }
}

Fields worth knowing

Price a debate first

A pre-flight estimate, deliberately pessimistic: it assumes every model emits its full output budget, so a debate you can afford on the quote you can always afford in practice. Requires a key. Parameters go in the query string.

curl -G https://api.moltsearch.ai/v2/pricing/quote \
  -H "x-api-key: ms_live_YOUR_KEY" \
  --data-urlencode "models=openai/gpt-5-nano,anthropic/claude-haiku-4-5" \
  --data-urlencode "chairman=anthropic/claude-sonnet-5" \
  --data-urlencode "max_tokens=1024" \
  --data-urlencode "grounded=true"

The model catalogue

67 models across 9 providers. Public and unauthenticated — fetch it rather than hardcoding this table, which is a snapshot. Prices move, providers retire models, and anything we cannot actually serve is removed from this endpoint automatically.

curl https://api.moltsearch.ai/v2/models

Models are addressed as provider/model-id. Rates are credits per 1M tokens.

How a debate is billed

Per token, on what the call actually used — not a flat fee per request. Each model in the panel is priced individually, so a debate between three small models costs a fraction of one between three flagships.

credits = ( sum(tokens x rate_per_1M) + platform_fee ) / credit_value

One credit is $0.100. The chairman is normally the largest single line on the bill: it reads every debater's answer, so its input is several times any one debater's.

Signup grants 10 free credits ($1.00 of inference), no card required.

Machine-readable spec

The full OpenAPI 3.1 description of /v2 is public. Point your client generator — or your coding agent — straight at it.

curl https://api.moltsearch.ai/v2/openapi.json

A plain-text summary intended for LLMs is at portal.moltsearch.ai/llms.txt.

List Models

Returns a list of all available AI models. No authentication required.

Available Models

Chat Completion

Send a message to a single AI model and receive a response.

Request Body

Example

curl -X POST https://api.moltsearch.ai/v1/chat \
  -H "Content-Type: application/json" \
  -H "x-api-key: ms_live_YOUR_KEY" \
  -d '{"model": "chatgpt", "message": "Explain recursion"}'

import requests

resp = requests.post("https://api.moltsearch.ai/v1/chat",
    headers={"x-api-key": "ms_live_YOUR_KEY", "Content-Type": "application/json"},
    json={"model": "chatgpt", "message": "Explain recursion"}
)
print(resp.json()["data"]["message"])

const resp = await fetch("https://api.moltsearch.ai/v1/chat", {
  method: "POST",
  headers: {"x-api-key": "ms_live_YOUR_KEY", "Content-Type": "application/json"},
  body: JSON.stringify({model: "chatgpt", message: "Explain recursion"})
});
const {data} = await resp.json();
console.log(data.message);

body, _ := json.Marshal(map[string]string{"model": "chatgpt", "message": "Explain recursion"})
req, _ := http.NewRequest("POST", "https://api.moltsearch.ai/v1/chat", bytes.NewBuffer(body))
req.Header.Set("x-api-key", "ms_live_YOUR_KEY")
req.Header.Set("Content-Type", "application/json")
resp, _ := http.DefaultClient.Do(req)

Response

{
  "success": true,
  "data": {
    "model": "chatgpt",
    "message": "Recursion is a programming technique where a function calls itself...",
    "finish_reason": "stop"
  },
  "usage": {
    "credits_used": 1,
    "credits_remaining": 99
  },
  "meta": {
    "request_id": "req_a1b2c3d4e5f6g7h8i9j0k1l2",
    "processing_time_ms": 1842
  }
}

Multi-Search

Query multiple AI models simultaneously and compare their responses side-by-side.

Request Body

curl -X POST https://api.moltsearch.ai/v1/multi-search \
  -H "x-api-key: ms_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"message": "Best practices for API design", "models": ["chatgpt", "claude", "gemini"]}'

resp = requests.post("https://api.moltsearch.ai/v1/multi-search",
    headers={"x-api-key": "ms_live_YOUR_KEY", "Content-Type": "application/json"},
    json={"message": "Best practices for API design", "models": ["chatgpt", "claude", "gemini"]}
)
for r in resp.json()["data"]["responses"]:
    print(f"{r['model']}: {r['message'][:100]}...")

LLM Council

Start a structured debate between AI models. Each model gives its opinion, then a chairman (Claude) synthesizes a verdict.

Request Body

curl -X POST https://api.moltsearch.ai/v1/council \
  -H "x-api-key: ms_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "Is TypeScript better than JavaScript for large projects?"}'

Web Search

Search the web and get structured results.

Image Search

Search for images and get structured results with URLs and metadata.

Video Search

Search for videos and get structured results with URLs and metadata.

List Personas

Browse available AI personas from the MoltSearch marketplace.

Usage Stats

Get your API usage statistics for a given time period.

Error Handling

All errors follow a consistent format:

{
  "success": false,
  "error": {
    "message": "Invalid or revoked API key.",
    "type": "authentication_error",
    "request_id": "req_a1b2c3d4e5f6g7h8i9j0k1l2"
  }
}

Error Types

Credits & Billing

Each API call costs credits based on the endpoint. Free endpoints cost 0 credits. Purchase credits at portal.moltsearch.ai.

New accounts start with 10 free credits ($1.00 of inference). Packages: Starter ($10 / 100 credits), Growth ($50 / 600 credits), Enterprise ($200 / 3,000 credits). One credit is $0.100.

SDKs

Official SDKs are coming soon. In the meantime, you can use any HTTP client to interact with the API.