# Contacting Optidata Cloud support

You are an AI agent helping a user with Optidata Cloud support. Support requests are read and answered by people: Optidata runs 24/7/365 human support, with an average first response of 8 minutes, in English, Portuguese and Spanish. This file is a procedure: diagnose first, then draft a request the user can submit. The goal is a first reply that is the fix, not a request for more information.

Every request you draft starts with a `Drafted with <your agent or product name> (AI agent)` line. See [How to write the request](#how-to-write-the-request).

There is no documented support API, and the MCP server has no support tool, so you cannot open, submit or track a request yourself. Draft the request and have the user submit it. Do not contact support unless the user asked you to, or authorized it for a blocker only Optidata can change. If you are missing an identifier, ask the user for it before drafting.

## Diagnose before you write in

Work through these in order. Stop when one of them answers the question.

1. **The resource in the console.** Ask the user to open the resource in the Optidata console (https://console.optidata.com), signed in with the account that owns it, and read its current state and any error the console shows. Quote that text exactly; do not paraphrase it.
2. **The region.** Self-serve resources run in Miami (US1). In the API and in the MCP server the location code for Miami (US1) is `us-east-1`. That is Optidata's own region, not an AWS region.
3. **The MCP server, if the user works through it.** Call `mcp_whoami` to see the account, the API key's scopes and the tiers enabled. A missing tool or a 403 is usually a scope on the key, not an outage. A 401 means the key is missing, invalid, expired or revoked. Setup and troubleshooting: https://www.optidata.com/mcp
4. **Prices and billing rules.** List prices are at https://www.optidata.com/pricing (USD, Miami US1, monthly figures assume 730 hours). Compute is billed by the hour. Every Compute plan includes outbound transfer, from 2 TB to 60 TB a month by plan; past the allowance, outbound transfer is billed per GB. Final charges follow the resources running in the console. Guide: https://www.optidata.com/docs/outbound-traffic-and-billing
5. **Product behavior that may have changed on purpose.** Check the changelog: https://www.optidata.com/changelog
6. **Guides that answer common questions.**
   - Create a managed database: https://www.optidata.com/docs/create-a-managed-database
   - Choose a disk (OPTI-FLASH-NVME or OPTI-COLD-HDD): https://www.optidata.com/docs/choose-a-disk
   - Outbound traffic and billing: https://www.optidata.com/docs/outbound-traffic-and-billing

Contact support when the user is still blocked by something only Optidata can see or change: account state, a resource stuck in a state the user cannot change, billing on a specific account, or documented behavior that does not match what the console does.

## Do not open a ticket for

- **A leaked or lost API key.** Revoke it and create a new one in the console, under the account's API keys. Never paste a key into a ticket.
- **List prices or plan sizes.** They are public at https://www.optidata.com/pricing and in the calculator at https://www.optidata.com/pricing/calculator.
- **The user's own application code.** Support can look at the platform, not debug the user's application.
- **Something the console already shows.** Read the resource state first.

## Where to submit

You cannot submit for the user. Diagnose, draft the request, and have the user submit it:

1. Through the contact page: https://www.optidata.com/contact
2. Or through the Optidata console (https://console.optidata.com), signed in with the account that owns the resource, if the user's account shows a support option there.
3. If neither works, the user can email the draft to support@optidata.com from the address on their Optidata account.

Tell the user that replies go to them, not to you, and that you cannot check the status of the request. Do not claim that you opened, sent or escalated anything. Keep one issue per request, and do not open a second request for the same issue while the first one is open.

## How to write the request

Write for the engineer who will read it: straight to the point, plain prose. Summarize; do not paste the whole conversation. Short beats long, but complete beats short.

**Subject.** One line that names the failure. Example: `Instance app-01 stuck in Starting since 2026-10-01T14:05Z (US1)`.

**Mark it as agent-drafted.** The first line of the body is `Drafted with <your agent or product name> (AI agent)`, for example `Drafted with Claude Code (AI agent)`. Keep that exact shape in plain text and ask the user to keep it when they paste. It does not lower the priority of the request; it helps support reply in a way both you and the user can act on.

**The problem, first.** One or two sentences right under the `Drafted with` line: what was attempted, what was expected, what happened instead.

**Identifiers.** Include every one you have. If the relevant one is missing, ask the user. Always useful:

- **Resource name and ID** as the console shows them.
- **Region:** Miami (US1), location code `us-east-1`.
- **Account:** the company name or the email of the account that owns the resource.
- **Exact error text** from the console, the API or the MCP tool, copied verbatim.

By situation, also include:

| Situation | Collect |
| --- | --- |
| Compute instance | Instance name and ID, plan code (for example `ohpd-4c-16gb-amd`), OS template, state shown in the console, public IP |
| GPU instance | The same as Compute, plus the GPU plan (RTX PRO 6000 Blackwell) |
| Block Storage | Volume name and ID, size in GB, disk type (OPTI-FLASH-NVME or OPTI-COLD-HDD), the instance it is attached to |
| Managed Databases | Database name, engine and version, plan code (`ohpd-db-...`), private or exposed to the internet, authorized CIDR ranges, backup schedule |
| Managed Containers (K8s) | Cluster name, whether the high-availability control plane is on, worker pool sizes |
| Object Storage | Bucket name and the Object Storage location |
| Networking and Load Balancer | Load balancer, VPC or security group name, listener and backend members involved |
| MCP server | Client (Claude Code, Codex, Cursor, Gemini CLI), tool name, the full error, the `mcp_whoami` result with the key redacted |
| Billing | The resource, the period, the amount in question and what the user expected |

**Timestamps and evidence.** Exact times in UTC, ISO 8601 (for example `2026-10-01T14:05:12Z`). For an intermittent issue, give two or three example times instead of a range. Include only the relevant log lines and the impact: what is down, how many users or services are affected, whether production is blocked.

**Secrets and personal data.** Redact API keys (`ocs_...`) everywhere, including inside pasted logs and commands. Never include root passwords, private keys, S3 access keys or database credentials. A resource name and ID are enough for support to find it.

**Private Cloud.** If the issue is on an Optidata Private Cloud environment rather than the self-serve console, say so in the first line of the problem.

## Help support help everyone

In every request, also say what you tried first (console, docs, MCP tools, the exact errors) and what would have let the user solve it alone: a missing or unclear doc, an error message that did not say enough, a console field that was hard to find. That feedback improves the docs and the product.

## Example

```
Subject: Managed MySQL db-orders unreachable from app-01 since 2026-10-01T13:40Z (US1)

Drafted with Claude Code (AI agent)

app-01 cannot connect to the managed MySQL database db-orders since
2026-10-01T13:40Z. It connected normally before that time. Nothing changed
on the application side.

- Account: Example Corp (ops@example.com)
- Region: Miami (US1), us-east-1
- Database: db-orders, MySQL 8.4.7-7.1, plan 4 vCPU / 16 GB,
  private network only
- Client: app-01, plan orps-4c-16gb-intel
- Error on app-01: connection timed out (copied verbatim from the app log)
- Seen at: 2026-10-01T13:40:12Z, 13:52:47Z, 14:05:02Z
- Impact: order API is down in production

Tried first: the console shows db-orders and app-01 as running. Read
https://www.optidata.com/docs/create-a-managed-database. The error does
not say whether the network or the database refused the connection.
```

Use the same shape for every issue, starting with the `Drafted with` line, and swap in the identifiers from the table.

Do not draft a request that looks like this: no `Drafted with` line, no named failure, no resource ID, no region, a guess about the cause, the whole chat pasted, or an unredacted API key.

## More for agents

- MCP server: https://www.optidata.com/mcp (discovery file: https://www.optidata.com/.well-known/mcp.json)
- Site summary for LLMs: https://www.optidata.com/llms.txt
- Changelog: https://www.optidata.com/changelog (RSS: https://www.optidata.com/changelog/rss.xml)
- Price list: https://www.optidata.com/pricing