# Revnu API, CLI and MCP Docs > The Revnu API, CLI and MCP server let your software read and approve the review queue, message Revnu, pull results and leads, and run workflows. Start here. Canonical: https://revnu.com/docs/developers Last updated: 2026-10-16 Revnu has a JSON API over HTTPS, a `revnu` command line tool, and an MCP server for Claude, ChatGPT, Cursor and other assistants. All three reach the same account through the same personal tokens, so anything you can do in one you can do in the others. They give your software the controls a customer has in the dashboard. Read what is waiting for approval and approve or decline it. Hand Revnu a job and read its reply. Pull results, leads and workflow runs into your own reporting. Keep a block list in sync. The API is Revnu's front door, not a toolbox. No route writes and sends an email of your own or publishes your own content. Everything that sends, publishes or spends is something Revnu drafted, approved through the same review cards and approval settings the dashboard uses. ## The contract in one screen ``` Base URL: https://revnu.com/api/v1 Auth: Authorization: Bearer rvp_... (server-side only, no CORS) Format: JSON in, JSON out, every response Cache-Control: no-store MCP: https://revnu.com/api/mcp (Streamable HTTP, OAuth) CLI: npx @revnu/cli (or npm install -g @revnu/cli) Spec: https://revnu.com/openapi.json ``` ## What you can do - **Review queue.** List what is waiting, read one card, approve, answer, choose, send, request changes, snooze, dismiss. See [Build a Review Screen](/docs/build-a-review-screen). - **Agent.** Send Revnu a message, wait for or poll the reply, read the thread, check whether it is busy. See [Talk to Your Agent](/docs/talk-to-your-agent). - **Results.** The same per-channel scoreboard Revnu reads when it wakes, for any window up to 31 days. See [Pull Results Into Reporting](/docs/pull-results-into-reporting). - **Leads.** Read leads, their contacts, one lead's draft and research. Set a lead's labels. Manage the block list. - **LinkedIn.** Look up one person's connection, invite and last 50 messages with your connected accounts. Read-only. - **Workflows.** Create, edit, pause, resume, delete and run recurring workflows, and read their run history. - **Setup.** Create and set up a new account from a terminal or an AI assistant. See [Install the CLI](/docs/cli). - **Preview.** Research a business's website with no account and no token, and hand the person a claim link. See [Preview a Plan Without an Account](/docs/preview). - **Inbound webhooks.** Give any system a secret URL; each JSON POST hands Revnu the payload. See [Inbound Webhooks](/docs/webhooks). ## What it will not do - No route sends an email you wrote or publishes a post you wrote. - No outbound webhooks or event subscriptions. Poll [results](/docs/results), [leads](/docs/leads) or the [review queue](/docs/review-list) instead. - No route creates leads in bulk. Hand Revnu a list as a [message](/docs/agent-messages) or a [webhook](/docs/webhooks) and it researches them. - No provider tools: nothing exposes the mailbox, LinkedIn, ad or search accounts directly, and there are no settings writes. The full comparison, including whether to build your own growth agent instead, is on [Build Your Own Growth Agent or Use Revnu?](/docs/api-cli-mcp-or-chat) ## Give your coding agent its own token If Claude Code, Codex, Cursor or another coding agent will talk to Revnu for you, give it its own token rather than sharing yours. Two ways to get one: - **`revnu login`** from the [CLI](/docs/cli). It opens a browser where you sign in and tick the permissions the token gets, then saves a token labelled "Revnu CLI", with the agent's name when an agent ran it. Connecting the [MCP server](/docs/mcp) does the same, labelled with the app's name. - **Settings → Developers → API tokens** in the dashboard. Name it after the agent, such as `claude-code-laptop`, and pass it as `REVNU_API_TOKEN`. Give it a token even if it can drive a browser. A coding agent that types into the dashboard chat is recorded as you: same name, all your permissions, no label, nothing to revoke on its own. Tick only what the agent needs. To hand Revnu jobs and check on them, that is `agent:message` plus `results:read`. Add `review:read` if it should report what is waiting for you, and `leads:read` if it reads leads. Leave out `review:write` unless you want the agent approving and sending on your behalf. The label is how you tell its work apart from yours. Revnu sees its messages as coming from `API token "