FAQ & Troubleshooting

Common questions and issues we've seen. If you're stuck, start here first.

Dashboard & product

Do I have to use MCP, or can I use the app only?

You can use either or both. The dashboard at app.adros.ai (opens in a new tab) drives the same backend as MCP and REST: Command Center, Strategy Workspace, Files, Campaigns, and CEO.

Where do approvals live?

In-app: Command Center (/command-center or /dashboard) and the Approvals page when wired.
REST: POST /api/v1/tasks/{task_id}/approve, POST /api/v1/tasks/{task_id}/reject, and GET /api/v1/approvals/pending.

Command Center says "failed" even though files were created

File presence is not the same thing as task completion. A task can upload artifacts and still end in review, paused, or failed if required deliverables are incomplete, package-level work is still outstanding, or the uploaded outputs need revision.

What to check next:

Open the task detail and read the latest timeline entries.
Check whether an approval, blocker, or follow-up step is still open.
If you are integrating via REST, compare GET /api/v1/tasks/{task_id} with GET /api/v1/tasks/{task_id}/events instead of using file count alone.

Why did Step 1 land as 5 files instead of 7?

Some research packages are staged. In many flows, the core written strategy/research docs land first, while supporting spreadsheet-style deliverables are handled as follow-up outputs in the same overall package.

Treat the package as complete only when Command Center marks it complete or your operator confirms completion. If you are integrating programmatically, key off task status and the deliverables expected for your package rather than a raw file count.

Quality track: scorecards or guardrails missing

Check these first:

The backend deploy you are calling includes the quality-track code
The target database has the required migration
The target environment has quality track enabled
The task is an audit-type task that actually completed the quality path

Connection & setup

404 on contract endpoints

You likely dropped the /api/v1 prefix.

Correct: https://api.adros.ai/api/v1
Wrong:   https://api.adros.ai

401 Unauthorized on every call

Common causes:

Wrong header name
Wrong token
Rotated token

REST uses Authorization: Bearer lyr_YOUR_TOKEN. MCP uses X-User-Token: lyr_YOUR_TOKEN.

Claude Desktop shows 0 tools after adding Adros

Adros uses Streamable HTTP, not stdio.

{
  "mcpServers": {
    "adros": {
      "url": "https://api.adros.ai/mcp",
      "headers": { "X-User-Token": "lyr_YOUR_TOKEN" }
    }
  }
}

Restart Claude Desktop fully after changing the config.

Cursor can't find Adros tools

Check that:

MCP is enabled in Cursor
The server type is HTTP, not stdio
You restarted Cursor after adding the config
Your token and URL are correct

Monitoring & webhooks

Daily webhook never fires even though I configured the URL

Check:

Your environment actually has monitoring enabled
Relevant ad accounts are connected
There may be no issues to report

Use POST /api/v1/me/monitoring/run-now to force a test run when supported by your deployment.

HMAC signature doesn't verify on my receiver

Use the raw request body and verify against X-Adros-Signature. Your receiver should be idempotent and should not rely on retries.

Creative generation

`generate_creative` returns the same image twice

There is an idempotency window to prevent duplicate results from retry storms. If you need a new variation, change the prompt or wait for the cache window to pass.

`generate_creative` returns 429 Too Many Requests

Creative generation is rate-limited. Wait for the window to reset or reduce duplicate calls.

Image looks generic or off-brand

Usually this is a prompt-construction issue, not a generation issue. Make sure you are using the correct workflow, brand context, and generation settings for your environment.

Contract Layer Dashboard ↗