Happytails.ai API | agent view

Your first request

Choose a callable tool ID and POST to /api/v1/tools/{id}/run. This example turns text into camelCase.

curl https://happytails.ai/api/v1/tools/case-converter/run \
  -H 'Content-Type: application/json' \
  -d '{"input":"Hello from an agent","options":{"mode":"camel"}}'

import json
from urllib.request import Request, urlopen

request = Request(
    "https://happytails.ai/api/v1/tools/case-converter/run",
    data=json.dumps({
        "input": "Hello from an agent",
        "options": {"mode": "camel"}
    }).encode(),
    headers={"Content-Type": "application/json"},
)
with urlopen(request, timeout=30) as response:
    print(json.load(response)["result"]["text"])

const response = await fetch(
  "https://happytails.ai/api/v1/tools/case-converter/run",
  {
    method: "POST",
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({
      input: "Hello from an agent",
      options: { mode: "camel" }
    })
  }
);
const data = await response.json();
if (!response.ok) throw new Error(data.error.message);
console.log(data.result.text);

Response · 200 OK

{
  "tool": "case-converter",
  "result": {
    "text": "helloFromAnAgent",
    "files": [],
    "name": "result.txt",
    "mime": "text/plain;charset=utf-8"
  }
}

Authentication

The local preview does not require an API key. A public endpoint, authentication scheme and account quotas have not been launched. Do not send requests to happytails.ai; configure your own API origin.

Browser clients must use the same origin. Cross-origin browser requests are not enabled in this preview. Server-side clients can call the deployed origin directly.

Discover tools and schemas

OpenAPI 3.1.1 describes every callable operation. The llms.txt file helps agents find the documentation. Not every catalog page has a callable API: browser-only and unconnected tools are omitted from discovery.

Request body

Unknown properties, unsupported options and invalid option values return an error. Check the tool's schema instead of inferring option names from its title.

Response body

Successful responses include tool and result. result.text is text or null; result.data contains parsed JSON when applicable; result.files is an array of downloadable results. Text results may also include name and mime. Preserve the returned MIME type and review the tool's fidelity notes before using a conversion.

Send and receive files

To create a ZIP containing a small text file, POST this body to /api/v1/tools/create-zip/run:

{
  "files": [
    {
      "name": "hello.txt",
      "mime": "text/plain",
      "base64": "SGVsbG8="
    }
  ]
}

Each file contains a name and padded base64 content. mime is optional. An optional bytes count must match the decoded content. Directory paths and file URLs are not accepted.

The response returns the generated archive in result.files[0], with name, mime, bytes and base64. Decode the base64 to save the ZIP; do not save the base64 text itself as the file. You can pass returned file objects to another compatible tool.

from base64 import b64decode
from pathlib import Path

# response_data is the parsed API response
file = response_data["result"]["files"][0]
Path("result.zip").write_bytes(b64decode(file["base64"]))

import base64
import json
from pathlib import Path
from urllib.request import Request, urlopen

source = Path("hello.txt")
request = Request(
    "https://happytails.ai/api/v1/tools/create-zip/run",
    data=json.dumps({"files": [{
        "name": source.name,
        "base64": base64.b64encode(source.read_bytes()).decode("ascii")
    }]}).encode(),
    headers={"Content-Type": "application/json"},
)
with urlopen(request, timeout=30) as response:
    archive = json.load(response)["result"]["files"][0]
Path("result.zip").write_bytes(base64.b64decode(archive["base64"]))

Files are processed in memory or disposable server working directories and returned in the response; the API does not provide a persistent download URL.

Limits

Errors

{
  "error": {
    "code": "INVALID_INPUT",
    "message": "mode is not an allowed value."
  }
}

Retries and timeouts

Use bounded exponential backoff with jitter for 429 responses. Correct 400, 413 and 415 requests before retrying. For a timeout or lost connection, the caller may not know whether processing finished. Retrying can repeat work or produce different random/model output. There is no idempotency-key contract, persistent job queue, polling endpoint or webhook API in this version. Execution is synchronous: the response contains the result.

Browser agents and WebMCP

The HTTP API works with scripts and agents without opening a tool page. In a browser that supports WebMCP, discover the tools exposed by the current page. Availability depends on the browser and the selected tool.

Legacy text pages expose transform_current_tool with {"text":"Hello world"}. It uses the visible settings and returns result and counts. Other supported pages expose run_current_tool with input and optional options. It returns result.text and result.download_count, or error.code and error.message. Omitted settings use the displayed defaults.

Select files on the page before calling a file tool. Read the displayed download links to retrieve generated files. Calls do not automatically copy or download. Browser permissions still apply. Network tools contact the supplied public target; tools marked as server processing upload their input to our backend. A page may support browser interaction without a callable HTTP endpoint. Use the tool capability manifest for availability.

Processing and privacy

API calls process your input on the server, even when the human version runs locally in a browser. Network diagnostics can contact public targets and providers. The application does not retain tool request bodies or generated files. Your client or agent platform may keep its own history. See the Privacy policy.

The local preview requires no authentication and binds to loopback only. It does not expose local filesystem paths. A remote or cloud agent cannot reach your laptop's 127.0.0.1 address.

Callable tool reference

Search by tool name, ID or category. Each API reference includes parameters, defaults, supported files, limits, processing notes and complete schemas.

437 tools