Use OpenAdManager from n8n, Make, Zapier and scripts
Your automation tool or script talks to the same OpenAdManager MCP server your AI app uses. You give it a personal token instead of signing in with a browser. Personal tokens are available on every plan, Free included.
| Field | Value |
|---|---|
| Server URL | https://mcp.openadmanager.com/mcp |
| Transport | Streamable HTTP |
| Authentication | Authorization: Bearer <your personal token> |
1. Make a personal token
- Open Add to your AI in the menu and click Connect to AI.
- Below the list of apps, choose Use a personal token.
- Give it a Name you will recognise, for example
n8n weekly report. - Under What it can do:
- See your ads and results is always on.
- Create and change ads is ticked for you. Untick it for reports and dashboards, and keep it only if the automation must change something. If your plan only allows reading, this choice is not shown. A personal token can never be given the spend permission.
- Under Works in, choose Only the ones I pick and tick only the ad accounts the automation needs.
- Pick Expires: 7 days, 30 days, 90 days or 1 year. Shorter is safer. You can make a new one at any time.
- Click Limit to IP addresses (optional) and add the address of your server or automation tool. The token will not work from anywhere else.
- Click Create token, then copy it. It is shown once. Put it in your tool's secret store or an environment variable, never in a workflow note, a public repository or a chat.
Each token is a row in the table on Add to your AI. Its menu has Settings (works in, IP addresses, expiry), Rename, Get a new token, Suspend and Revoke. Revoke it as soon as you stop using it.
Only owners, admins and members can create tokens. Viewers cannot.
2. Connect your tool
n8n
Add an MCP Client Tool node to your AI Agent and set:
| n8n field | Value |
|---|---|
| Endpoint | https://mcp.openadmanager.com/mcp |
| Server Transport | HTTP Streamable |
| Authentication | Bearer Auth |
| Credentials | A new Bearer credential with your personal token |
Use Tools to Include > Selected to give the agent only the tools it needs. Older n8n versions only have a field called SSE Endpoint and no transport choice. Update n8n, because this server uses Streamable HTTP.
Checked against the n8n MCP Client Tool node (version 1.2 and later).
Make
Add the MCP Client module to a scenario. Enter https://mcp.openadmanager.com/mcp as the server URL, choose token-based authentication instead of OAuth, and paste your personal token when Make asks for it. Make loads the tool list from the server, and you pick a tool and map its inputs.
Zapier
Zapier's MCP Client (a beta feature) is added as an app connection: Apps > Add connection > MCP Client.
| Zapier field | Value |
|---|---|
| Server URL | https://mcp.openadmanager.com/mcp |
| Transport | Streamable HTTP |
| OAuth | No |
| Bearer Token | Your personal token |
In a Zap use Run Read-Only Tool for reports, and Run Tool only for something that changes ads. Zapier's MCP Client only supports tool calls.
3. Call it from a script
Set the token as an environment variable first:
export OAM_TOKEN='paste-your-token-here'
The server replies with one server-sent event, so the response looks like event: message followed by a data: line that holds the JSON. Always send Accept: application/json, text/event-stream. Without it the server answers 406 Not Acceptable. No session is needed: each call stands alone.
curl
Say hello:
curl -s https://mcp.openadmanager.com/mcp \
-H "Authorization: Bearer $OAM_TOKEN" \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"my-script","version":"1.0"}}}'
List the tools (the same long list your AI app sees):
curl -s https://mcp.openadmanager.com/mcp \
-H "Authorization: Bearer $OAM_TOKEN" \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","id":2,"method":"tools/list"}'
List your ad accounts:
curl -s https://mcp.openadmanager.com/mcp \
-H "Authorization: Bearer $OAM_TOKEN" \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"oam_list_accounts","arguments":{}}}'
The reply (values changed):
event: message
data: {"result":{"content":[{"type":"text","text":"{\"accounts\":[...]}"}],"structuredContent":{"accounts":[{"platform":"meta","account_id":"act_1234567890","name":"Example shop","client":"Main","currency":"USD","connection":{"label":"Meta sign-in","status":"ok"}}]},"isError":false},"jsonrpc":"2.0","id":3}
Read result.structuredContent for data. tools/list returns result.tools, each with name, title, description and inputSchema.
JavaScript (Node 18 or newer)
const URL = "https://mcp.openadmanager.com/mcp";
async function rpc(method, params = {}) {
const res = await fetch(URL, {
method: "POST",
headers: {
Authorization: `Bearer ${process.env.OAM_TOKEN}`,
"Content-Type": "application/json",
Accept: "application/json, text/event-stream",
},
body: JSON.stringify({ jsonrpc: "2.0", id: 1, method, params }),
});
if (!res.ok) throw new Error(`HTTP ${res.status}`);
const text = await res.text();
// The reply is either plain JSON or one server-sent event ("data: {...}").
const line = text.split("\n").find((l) => l.startsWith("data:"));
const msg = JSON.parse(line ? line.slice(5) : text);
if (msg.error) throw new Error(msg.error.message);
return msg.result;
}
const tools = await rpc("tools/list");
console.log(tools.tools.length, "tools, first:", tools.tools[0].name);
const out = await rpc("tools/call", { name: "oam_list_accounts", arguments: {} });
for (const a of out.structuredContent.accounts) console.log(a.platform, a.account_id, a.name);
Save it as accounts.mjs and run node accounts.mjs. Output:
304 tools, first: google_ads_field_metadata
meta act_1234567890 Example shop
google 1234567890 Example ads
The tool count grows over time.
Python 3 (no packages)
import json, os, urllib.request
URL = "https://mcp.openadmanager.com/mcp"
def rpc(method, params=None):
req = urllib.request.Request(
URL,
data=json.dumps({"jsonrpc": "2.0", "id": 1, "method": method, "params": params or {}}).encode(),
headers={
"Authorization": "Bearer " + os.environ["OAM_TOKEN"],
"Content-Type": "application/json",
"Accept": "application/json, text/event-stream",
},
)
with urllib.request.urlopen(req) as res:
text = res.read().decode()
# The reply is either plain JSON or one server-sent event ("data: {...}").
line = next((l for l in text.splitlines() if l.startswith("data:")), None)
msg = json.loads(line[5:] if line else text)
if "error" in msg:
raise RuntimeError(msg["error"]["message"])
return msg["result"]
tools = rpc("tools/list")["tools"]
print(len(tools), "tools, first:", tools[0]["name"])
out = rpc("tools/call", {"name": "oam_list_accounts", "arguments": {}})
for a in out["structuredContent"]["accounts"]:
print(a["platform"], a["account_id"], a["name"])
The output has the same shape as the JavaScript one.
What a wrong call looks like
| You did | You get |
|---|---|
| Wrong, expired or revoked token | HTTP 401 with a WWW-Authenticate header. Make a new token. |
Left out the Accept header | HTTP 406 Not Acceptable. Send both content types. |
| Called a change tool with a read-only token | A normal reply with isError: true and the code OAM_INSUFFICIENT_SCOPE. Nothing was changed. |
| Called a change that is switched off in AI permissions | A reply that says nothing was changed and names the switch. See AI permissions. |
| Went over your plan | A reply with isError: true and a code such as OAM_QUOTA_CALLS. The message says what ran out and when it resets. |
Limits
A token counts against the same MCP requests and the same speed limits as your AI apps, on the same plan. Nothing extra is metered and nothing is faster for a script. See the limits table in the MCP reference and the plan details on Pricing. Changes follow your AI permissions exactly as they do for an AI app, and every change is in the activity log with the token's name.
Please do not
- Resell or share it. A token is for your own ad accounts, or the accounts of clients you manage. It is not a service for other businesses.
- Poll in a loop. Ask once for a report and keep the result. Run scheduled jobs a few times a day, not every few seconds. Constant polling uses up your MCP requests and trips the speed limit.
- Put the token where others can read it. If it leaks, revoke it at once and make a new one.
- Give it more than it needs. Read-only, the fewest ad accounts, an expiry and an IP address cover almost every report.
This is the OpenAdManager MCP server. It is meant for AI apps and for automations that call tools the same way. It is not a general programming interface to your ad platforms.
Related: Add it to your AI app · AI permissions · MCP reference · Security
Your next campaign is one sentence away.
Get notified