# Connect your agent to Eevo

> Add Eevo’s MCP server to Claude, ChatGPT, Claude Code, Codex or any MCP client, and let your agent ask Eevo.

> [!NOTE]
> Eevo is made to be used by your agent, as part of its work. Agents make many small decisions for people. Connected to Eevo, an agent asks it when a choice affects people, and shows you where their interests conflict. Every client below signs in with OAuth, so you need no API key. New accounts start with 5 Default mode decisions, each revisable up to 10 times, and 50 Direct mode decisions, free.

Pick the client you use. Each section is complete on its own.

## Claude

1. In Claude, open **Settings › Connectors** and choose **Add custom connector**.
2. Name it Eevo and enter the URL `https://eevo.io/api/arvo/mcp`.
3. Click **Add**, then **Connect**, and keep the default options. Sign in to Eevo, or create an account in the same step.
4. Ask Claude about a decision you face, and tell it to ask Eevo.

On Team and Enterprise plans, an Owner adds the connector once under **Organization settings › Connectors**.

## ChatGPT

1. In ChatGPT, open **Settings › Apps & Connectors › Advanced settings** and turn on **Developer mode**.
2. In **Apps & Connectors**, choose **Create**. Name it Eevo and enter the URL `https://eevo.io/api/arvo/mcp`.
3. Set authentication to **OAuth**, keep the other defaults and click **Create**. Sign in to Eevo, or create an account in the same step.
4. In a new chat, turn on Eevo, describe a decision you face, and tell ChatGPT to ask Eevo.

Custom connectors need ChatGPT Plus, Pro, Business, Enterprise or Edu.

## Claude Code

1. Add Eevo’s MCP server:

```shell
claude mcp add --transport http eevo https://eevo.io/api/arvo/mcp
```

2. In Claude Code, run `/mcp`, choose **eevo** and **Authenticate**. Sign in to Eevo, or create an account in the same step.
3. Let Claude Code work. When a choice affects people, it asks Eevo, or you tell it to.

For scripts and CI, send an API key instead of signing in: add `--header "Authorization: Bearer $EEVO_API_KEY"` to the first command.

## Codex

1. Add Eevo’s MCP server:

```shell
codex mcp add eevo --url https://eevo.io/api/arvo/mcp
```

2. Sign in to Eevo, or create an account in the same step:

```shell
codex mcp login eevo
```

3. Let Codex work. When a choice affects people, it asks Eevo, or you tell it to.

For scripts and CI, send an API key instead: add `--bearer-token-env-var EEVO_API_KEY` to the first command and skip the second.

## Any other MCP client

1. Add a remote MCP server named Eevo, with streamable HTTP transport and the URL `https://eevo.io/api/arvo/mcp`:

```json
{
  "mcpServers": {
    "eevo": {
      "type": "http",
      "url": "https://eevo.io/api/arvo/mcp"
    }
  }
}
```

2. If your client supports OAuth, keep it on and sign in to Eevo when it asks. If not, send an API key from the [Eevo console](https://eevo.io/keys) as a bearer token ([MCP server](https://eevo.io/docs/reference/mcp)).

## Ask

Give your agent a decision:

> We found a bug that shows some customers a wrong invoice total. A fix is ready, but the logs that would confirm it arrive in 40 minutes. Should we ship now or wait?

The agent calls `eevo_decide` with the customers, your team and your company as stakeholders, and shows one row per stakeholder. In clients that show MCP Apps it then shows the final decision as a card with `eevo_show_decision`. The decision is saved in your decision log in the console.

The MCP server's tool descriptions tell the agent when to call Eevo and how to present the answer, so it needs nothing else. To let it read these docs too, point it at [llms.txt](https://eevo.io/api/docs/llms.txt).

## Next

- [Ask Eevo via curl (Native API)](https://eevo.io/docs/tutorials/first-decision): call the API yourself.
- [Revise a decision](https://eevo.io/docs/how-to/revise): when the result shows the agent left something out.