One agent, one verdict. PromptRat runs a single AI agent instead of a room full of them - every call comes with the numbers behind it. How it works

Open data

Public API

The same machine-readable data behind the agent, free and keyless for any tool or bot. No signup, no wallet, no token needed.

Base URLhttps://promptrat.fun
AuthNo key · no signup · nothing to configure
Rate limit10 requests / min per IP
CORSOpen (Access-Control-Allow-Origin: *)
FormatJSON wrapped as { data, meta }
ErrorsOne shape for every error: { error: { code, message } }

Solana network health

This backend reads live from public Solana RPC nodes - epoch info, performance samples, vote accounts, cluster nodes and a sampled block. No indexer, no stale snapshot of the network.

solana-rpcloading…

Send a live request

Choose an endpoint, fill in the parameters, press Send - the answer comes back from the live API. Everything on this page is read-only and free.

Endpoints

The agent as a JSON card, filterable and paginated.

GET/api/agents
curl
curl https://promptrat.fun/api/agents
Response
Choose an endpoint and press Send to see the live JSON.

Response shape

Every 2xx reply shares one shape: data holds the payload, meta carries the version, timestamp, counts and the rate limit.

{ "data": [ { "id": "rat", "name": "RAT", "role": "Resident analyst - one agent, every read", "capabilities": ["hype", "momentum", "entry"] } ], "meta": { "version": "1.0", "generated_at": "2026-09-07T00:00:00.000Z", "rate_limit": { "limit": 10, "window_seconds": 60 }, "count": 1, "total": 10, "offset": 0, "limit": 1, "capability": "rug" } }

Errors keep one shape as well: a JSON body holding an error object. 404 means the resource is not there, 405 that the verb is not served (the Allow header names the ones that are), and 429 means you hit the rate limit (honour the Retry-After header).

{ "error": { "code": "NOT_FOUND", "message": "Agent 'nope' not found." } } HTTP 429 { "error": { "code": "RATE_LIMITED", "message": "Rate limit exceeded. Try again in 45s." } }

Code samples

No SDK and no key needed - a plain HTTP request does the job. Below is the same call in a few languages.

# curl # the agent as a JSON card curl https://promptrat.fun/api/agents # filter the agent by what it is good at, paginated curl "https://promptrat.fun/api/agents?capability=rug&limit=5&offset=0" # the agent's capability manifest curl https://promptrat.fun/api/agents/rat/capabilities
# Node.js // no dependencies - just fetch const res = await fetch("https://promptrat.fun/api/agents"); const { data, meta } = await res.json(); console.log(data.map(a => a.name)); // ["RAT"] console.log(meta.rate_limit); // { limit: 10, window_seconds: 60 }
# Python import json, urllib.request with urllib.request.urlopen("https://promptrat.fun/api/capabilities") as r: body = json.load(r) print(body["data"]) # [{ "slug": "rug", "label": "Rug checks", "count": 3 }, ...]

Worth knowing

  • Read-only & free. Read-only endpoints: no writes, no auth, nothing to pay.
  • Rate limited. 10 requests per minute per IP, which keeps the service fast and fair for everyone.
  • Iterate politely. The API reuses the same shared snapshot as the site, so requests are cheap - still, back off a little on 429 instead of hammering.
  • Browser or server. CORS is wide open (*), so any front-end, serverless function, Node bot or plain script can call it.

Want the full story on the platform, the agent and the data sources? Read the product docs.

⚠️ Not financial advice

PromptRat is an entertainment and research tool. Memecoins are extremely volatile and most lose value. Nothing here is a recommendation to buy or sell.