# Use Revnu From Your Coding Agent

> Connect Revnu to Claude Code, Codex, Gemini CLI or Cursor so your coding agent can hand Revnu growth jobs, report what's waiting for you, and read results, under its own token.

Canonical: https://revnu.com/docs/coding-agent
Last updated: 2026-10-16

Your coding agent already lives in your terminal and your repo. Connect Revnu to it and it can hand Revnu the growth work it can't do itself, tell you what's waiting for your approval, and pull numbers into whatever it's building.

## Install

```bash
curl -fsSL revnu.com/install | sh
```

This connects Revnu's MCP server to Claude Code, Codex and Gemini CLI, whichever you have, and adds the Revnu skill that tells your coding agent when Revnu can help. Nothing is created or paid for.

To connect one app by hand, or Cursor and VS Code, use the commands in [MCP Server Setup](/docs/mcp) or the install links on [revnu.com/agents](/agents).

## Sign in

Each app signs in once, in your browser:

### Claude Code

Run `/mcp` inside Claude Code, pick **revnu** and choose **Authenticate**.

### Codex

```bash
codex mcp login revnu
```

### Gemini CLI

Run `/mcp auth revnu` inside Gemini CLI.

No Revnu account yet? The same sign-in creates one, and your agent can take you through setup. See [Sign Up From an Agent](/docs/agent-signup).

## Ask it things

Your agent hands Revnu a job with the `ask_revnu` tool and reads the reply. Ask in plain words:

> Ask Revnu to find 50 seed-stage fintech founders in London and draft a first email to each.

> What's waiting for my approval in Revnu?

> Pull last week's Revnu results and add them to the weekly report.

> We just shipped the new pricing page. Tell Revnu so it can mention it in outreach.

Revnu does the work in your account and drafts land in your [review queue](/docs/review-list), the same as when you ask in the dashboard.

## Pick its permissions

When you sign in, you tick what the agent may do. Tick only what it needs:

| Permission | Lets the agent |
| --- | --- |
| Agent: Message (`agent:message`) | Hand Revnu jobs and read its replies |
| Results: Read (`results:read`) | Read the per-channel results |
| Review: Read (`review:read`) | Report what's waiting for you |
| Leads: Read (`leads:read`) | Read leads and their contacts |
| Review: Write (`review:write`) | Approve and send. Leave it off unless you want the agent sending for you |

Scopes are fixed once the connection is made. To change them, revoke it in **Settings → Developers** and connect again. Every scope is listed in [Authentication and Scopes](/docs/authentication).

## Its own token, not yours

Connecting creates a token labelled with the app's name, so Revnu sees the agent's messages as coming from that label and logs its review actions against it. To cut it off, revoke that one token in **Settings → Developers**. Nothing else stops working.

Don't let your agent drive the dashboard in a browser instead. Anything typed into the dashboard chat is recorded as you, with all your permissions and nothing to revoke on its own.

**Scripts, CI and the CLI**

For a script or a CI job, mint a token in **Settings → Developers**, name it after the job, and set it as `REVNU_API_TOKEN`. Or run `revnu login`, which opens the same sign-in and saves a token labelled "Revnu CLI".

```bash
revnu agent message "Draft a launch post for the new pricing page"
revnu review list
```

Without a token, CLI commands exit with code 3 and `NOT_AUTHENTICATED`. See [Install the CLI](/docs/cli).

## Add the skill by itself

The install script adds it already. To add it on its own:

```bash
npx skills add revnu-app/revnu-plugin
```

The skill tells your agent when a task is growth work for Revnu, such as finding customers, outreach or ads, rather than code to write.

## What it can't do

- It can't write and send an email of its own or publish its own post. It can only approve what Revnu drafted, and only with Review: Write.
- It can't reach your mailbox, LinkedIn, ad or search accounts directly. Revnu runs those.

Working on AI visibility? The fixes Revnu suggests are often site changes your coding agent can make. See [AI Visibility](/docs/ai-visibility).

## Frequently asked questions

### Does my coding agent need its own Revnu account?

No. It connects to your account with its own token, so its work shows up in your dashboard under its label.

### Can my coding agent approve and send?

Yes, when you tick Review: Write as it connects. Without it, the agent reads the review queue and leaves approvals to you.

### Why doesn't my agent have ask_revnu?

Either the connection didn't get Agent: Message, or your account hasn't started its trial yet. Reconnect and tick Agent: Message, or finish setup first.

### Does it work in Cursor?

Yes. Use the Cursor install link on [revnu.com/agents](/agents), or add `https://revnu.com/api/mcp` as a remote MCP server.

### Can I use Revnu from CI?

Yes. Set `REVNU_API_TOKEN` to a token minted for that job and call the [CLI](/docs/cli) or the [API](/docs/quickstart).