Magisterium AI

Developers
Share:

Getting Started

This guide walks you through making your first request to the Magisterium A2A endpoint, retrieving the resulting task, and reading the structured response.

Prerequisites

  1. A paid Magisterium plan (Pro, Organization, or Enterprise). Free accounts will receive a PLAN_REQUIRED error from A2A.
  2. An OAuth 2.0 access token tied to your magisterium.com account. A2A uses the same OAuth flow as the Magisterium MCP server; discover the authorization, token, and dynamic-registration endpoints from the OAuth metadata at https://www.magisterium.com/.well-known/oauth-authorization-server.
  3. The skills you want to call — see Skills for the full list.

Heads up: the long-lived API keys created in the API Console work for Chat Completions, Search, and News, but they are not valid for A2A — A2A only accepts OAuth-issued user tokens.

Export your access token as an environment variable so the examples below pick it up automatically:

bash
export MAGISTERIUM_TOKEN=<your-access-token-here>

Step 1: Discover the Agent

Before calling the endpoint, you can fetch the public Agent Card to see which skills are available and confirm the protocol version:

bash
curl https://www.magisterium.com/.well-known/agent.json

The card requires no authentication. Orchestrators can cache it for an hour — the endpoint sets Cache-Control: public, max-age=3600.

Step 2: Send Your First Message

All A2A operations go through a single JSON-RPC endpoint: POST https://www.magisterium.com/api/v1/a2a. The example below calls message/send on the catholic_qa skill.

bash
curl -X POST https://www.magisterium.com/api/v1/a2a \
    -H "Authorization: Bearer $MAGISTERIUM_TOKEN" \
    -H "Content-Type: application/json" \
    -d '{
      "jsonrpc": "2.0",
      "id": 1,
      "method": "message/send",
      "params": {
        "message": {
          "role": "user",
          "messageId": "msg-001",
          "kind": "message",
          "parts": [{ "kind": "text", "text": "What does the Church teach about the Real Presence?" }],
          "metadata": { "skillId": "catholic_qa" }
        }
      }
    }'
typescript
const accessToken = process.env.MAGISTERIUM_TOKEN!;

const response = await fetch("https://www.magisterium.com/api/v1/a2a", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${accessToken}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    jsonrpc: "2.0",
    id: 1,
    method: "message/send",
    params: {
      message: {
        role: "user",
        messageId: crypto.randomUUID(),
        kind: "message",
        parts: [
          { kind: "text", text: "What does the Church teach about the Real Presence?" },
        ],
        metadata: { skillId: "catholic_qa" },
      },
    },
  }),
});

const { result: task } = await response.json();
console.log(task.id, task.status.state);
for (const artifact of task.artifacts ?? []) {
  for (const part of artifact.parts) {
    if (part.kind === "text") console.log(part.text);
  }
}
python
import os
import uuid
import httpx

access_token = os.environ["MAGISTERIUM_TOKEN"]

payload = {
    "jsonrpc": "2.0",
    "id": 1,
    "method": "message/send",
    "params": {
        "message": {
            "role": "user",
            "messageId": str(uuid.uuid4()),
            "kind": "message",
            "parts": [
                {"kind": "text", "text": "What does the Church teach about the Real Presence?"}
            ],
            "metadata": {"skillId": "catholic_qa"},
        }
    },
}

response = httpx.post(
    "https://www.magisterium.com/api/v1/a2a",
    headers={
        "Authorization": f"Bearer {access_token}",
        "Content-Type": "application/json",
    },
    json=payload,
    timeout=60,
)
task = response.json()["result"]
print(task["id"], task["status"]["state"])
for artifact in task.get("artifacts", []):
    for part in artifact["parts"]:
        if part["kind"] == "text":
            print(part["text"])

Step 3: Read the Response

message/send returns a Task object with kind: "task". A synchronous skill call always ends in one of two terminal states:

  • status.state == "completed" — results live in result.artifacts. Each artifact has an artifactId, an optional name, and one or more parts (text, data, or file).
  • status.state == "failed" — the failure reason is in status.message.parts[0].text.
json
{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "id": "task_abc123",
    "contextId": "ctx_def456",
    "kind": "task",
    "status": { "state": "completed", "timestamp": "2026-04-20T12:00:00.000Z" },
    "artifacts": [
      {
        "artifactId": "art_ghi789",
        "name": "catholic_qa_response",
        "parts": [
          { "kind": "text", "text": "The Catholic Church teaches..." },
          { "kind": "data", "data": { "citations": [] } }
        ]
      }
    ]
  }
}

Step 4: Retrieve or Cancel a Task

Tasks are stored for 24 hours and can be refetched by ID using tasks/get:

bash
curl -X POST https://www.magisterium.com/api/v1/a2a \
    -H "Authorization: Bearer $MAGISTERIUM_TOKEN" \
    -H "Content-Type: application/json" \
    -d '{
      "jsonrpc": "2.0",
      "id": 2,
      "method": "tasks/get",
      "params": { "id": "task_abc123" }
    }'

If you need to abort a task, call tasks/cancel with the same ID. Because all Magisterium skills resolve synchronously, cancelation is only useful in rare retry scenarios — a task that has already completed or failed cannot be canceled and will return INVALID_PARAMS.

Common Errors

CodeMeaningFix
-32004UnauthorizedCheck the Authorization header and that the access token is valid and not expired.
-32005Paid plan requiredUpgrade your account at magisterium.com/plan.
-32002Skill not foundVerify metadata.skillId matches an ID listed in the Agent Card.
-32003Rate limit exceededWait retryAfter seconds (present in the error data) before retrying.

See the full API Reference for every method and error code.