# Connect AgentSEO

Endpoint: https://www.agentseo.dev/mcp

Documentation reviewed: 2026-11-14

## Claude Code — project setup

Use a server-side AgentSEO API key with no allowed domains. Analysis tools can consume credits and create jobs; repeated calls can incur another charge.

### Setup

#### Create a server-side API key

Keep the key private and supply it through AGENTSEO\_API\_KEY in the environment inherited by your client. Do not paste it into shared project configuration.

[Open AgentSEO API keys](https://www.agentseo.dev/dashboard/keys)

Sources: https://raw.githubusercontent.com/AgentSEO-dev/agentseo-clients/main/packages/mcp-server/README.md, https://www.agentseo.dev/docs/api-reference

#### Add the server to your client

Set AGENTSEO\_API\_KEY privately in your client environment. The example contains an environment reference, never a credential.

```json
{
  "mcpServers": {
    "agentseo": {
      "type": "http",
      "url": "https://www.agentseo.dev/mcp",
      "headers": {
        "Authorization": "Bearer ${AGENTSEO_API_KEY}"
      }
    }
  }
}
```

Sources: https://code.claude.com/docs/en/mcp

#### Share configuration, not credentials

Commit the project configuration after reviewing it. Each teammate supplies their own credentials or authorization and approves the project server in their client.

Sources: https://code.claude.com/docs/en/mcp

#### Inspect the available tools first

Connect and inspect the tool list. Choose a small task only after reviewing its credit cost and inputs. No paid analysis call was performed for this guide.

Sources: https://raw.githubusercontent.com/AgentSEO-dev/agentseo-clients/main/packages/mcp-server/README.md

#### If connection fails

Check that your client inherited AGENTSEO\_API\_KEY and the key is a server-side key with no allowed domains. Review account credits before running analysis.

Sources: https://raw.githubusercontent.com/AgentSEO-dev/agentseo-clients/main/packages/mcp-server/README.md

## Verification

- documentation: reviewed — Official setup documentation reviewed for these selected combinations. Runtime observations below are separate.
- endpoint: not-tested
- authentication: not-tested
- toolCall: not-tested
