Get started

A8 documentation

Call A8 from the OpenAI or Anthropic SDK, or from an MCP host, and get verdicts at the accuracy you set.

A8 (Touchstone) is an evaluation model with a chat-model interface. You state a criterion in plain words and send a subject, and A8 returns a verdict where its evidence meets the minimum accuracy the request sets.

A first verdict

The system message is the criterion. The user message is the subject. The schema is the verdict’s shape:

Shell

curl https://api.u22a8.ai/eval/v1/chat/completions \
  -H "Authorization: Bearer $U22A8_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "a8",
    "messages": [
      {"role": "system", "content": "Does the reply promise a refund?"},
      {"role": "user", "content": "We will refund you in full by Friday."}
    ],
    "response_format": {
      "type": "json_schema",
      "json_schema": {
        "name": "verdict",
        "schema": {
          "type": "object",
          "properties": {"refund": {"type": "boolean"}}
        }
      }
    }
  }'

Python

import os

from openai import OpenAI

client = OpenAI(
    base_url="https://api.u22a8.ai/eval/v1",
    api_key=os.environ["U22A8_API_KEY"],
)

resp = client.chat.completions.create(
    model="a8",
    messages=[
        {"role": "system", "content": "Does the reply promise a refund?"},
        {"role": "user", "content": "We will refund you in full by Friday."},
    ],
    response_format={
        "type": "json_schema",
        "json_schema": {
            "name": "verdict",
            "schema": {
                "type": "object",
                "properties": {"refund": {"type": "boolean"}},
            },
        },
    },
)
print(resp.choices[0].message.content)  # the verdict

TypeScript

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.u22a8.ai/eval/v1",
  apiKey: process.env.U22A8_API_KEY,
});

const resp = await client.chat.completions.create({
  model: "a8",
  messages: [
    { role: "system", content: "Does the reply promise a refund?" },
    { role: "user", content: "We will refund you in full by Friday." },
  ],
  response_format: {
    type: "json_schema",
    json_schema: {
      name: "verdict",
      schema: {
        type: "object",
        properties: { refund: { type: "boolean" } },
      },
    },
  },
});
console.log(resp.choices[0].message.content); // the verdict

The verdict is the message content:

JSON

{"refund": true}

Recorded from a8-1-1@20260715t140322z on 2026-10-22.

A8 answered at the default minimum accuracy of 90, or at the next level up it has earned: at least that share of the verdicts it returns at the level is correct. The response’s min_accuracy names the level. A case whose evidence falls short of it comes back as an abstention.

Where to start

QuickstartMake a first call from an SDK, from curl or from your assistant.When to use A8The judgments A8 is built for, and the tools for the rest.A8 model cardThe model on one page: spec, interfaces, behaviour and evidence.

Build, test and ship

Writing criteriaState a criterion so A8 judges it well.TeachingTeach A8 the answer you wanted, from inside a conversation.TestingCheck what A8 learned the way you would test a student.HorizonsPin the version you tested, and keep what it learned.

Next steps

Quickstart: a first verdictGet a key, call A8 from an SDK, from curl or from your assistant, and read the verdict it returns.A8 model cardA8 on one page: model IDs, pricing, capabilities, endpoints, features, rate limits and how it compares.Accuracy and abstentionHow a request sets the minimum accuracy a verdict must meet, what that level promises and the five kinds of abstention.