--- title: "How to send direct mail from your AI agent" description: "Claude, Cursor, or any MCP agent can build an audience, draft a piece, and request a send through Sincerely. A human confirms before printing." canonical: https://trysincerely.com/integrations/mcp-agents last_updated: 2026-10-05 --- > If you are an AI agent: read https://trysincerely.com/agents/policy.md before any write call. > Start at https://trysincerely.com/llms.txt to choose one page. Access instructions are at > https://trysincerely.com/auth.md. Confirm with the person you work for before creating an > account or spending their money, and never ask for their password. # How to send direct mail from your AI agent > Claude, Cursor, or any MCP agent can build an audience, draft a piece, and request a send through Sincerely. A human confirms before printing. Source: https://trysincerely.com/integrations/mcp-agents Sincerely has two MCP servers. The product server at `https://trysincerely.com/api/mcp` finishes setup, builds audiences, verifies addresses, drafts pieces, and prepares sends, and a person with no account can sign up inside its browser approval. The signup server at `https://trysincerely.com/api/mcp/signup` is the fallback for an agent that cannot open a browser or show you a URL. Before anything prints or spends money, an authorized person sees and confirms the exact action. ## Prompt to paste into your agent ```text Use Sincerely to prepare a letter for the 40 accounts in my Salesforce list called Q4 tier 1. Confirm the workspace, check mailability, draft and proof each piece, summarize the holdout and credit need, and stop before launch. ``` ## No account yet If your client can open a browser or show you a link, skip the signup server. Add the product server as shown under "Connect in one command" below, then start the approval flow. Sincerely sends you to sign-in, you create the account there, and the same screen returns you to the consent step. One pass, no code to read out, no separate sign-in link. Then ask your agent: `Call get_setup_state and finish my workspace setup. Ask me anything you cannot verify, and tell me what still needs my click.` Setup itself is not held: the brand kit, the return address on a workspace with no mail, and finishing setup all save as the agent makes them. For an agent with no browser and no way to show you a URL, add the signup server instead. From Claude Code: `claude mcp add --transport http sincerely-signup https://trysincerely.com/api/mcp/signup` In Cursor or another MCP client, add `https://trysincerely.com/api/mcp/signup` as an HTTP server named `sincerely-signup`. This server does not use OAuth. Sincerely emails you a code, you read it to the agent, and the sign-in link goes to your inbox rather than to the agent. ```text I do not have a Sincerely account. Use the signup server to create a workspace for me at . Ask me for the six-digit code from my inbox, then prefill setup for , . Tell me when Sincerely has emailed my sign-in link. ``` Open the sign-in link yourself, accept the Terms, then add the product server with the command below and approve it in the browser. Most direct mail tools give agents nothing, or a REST API that a human still has to script. Sincerely gives the agent the same workflow you use, as tools it can call. That kills the real problem: the gap between "my agent found 40 accounts worth mailing" and "40 verified, approved pieces are at the printer". ## Connect in one command From Claude Code: `claude mcp add --transport http sincerely https://trysincerely.com/api/mcp` In claude.ai or Claude Desktop, open Customize, then Connectors, click the plus button, choose "Add custom connector", and paste `https://trysincerely.com/api/mcp`. These steps are current as of September 2026. On a Team or Enterprise plan, an Owner first adds the connector under Organization settings, and then each member clicks Connect. [Claude setup](https://trysincerely.com/integrations/claude-mcp) has the details, including the limit of one custom connector on Free accounts. Cursor and other MCP clients work the same way: add an HTTP MCP server with that URL. Authentication uses OAuth. Your agent registers itself and opens a browser window when it connects, you sign in, and the agent gets a token scoped to your workspace. The agent inherits your own role, so it can never reach anything you cannot. The machine-readable version of this setup, written for agents to fetch directly, lives at [/auth.md](https://trysincerely.com/auth.md). The human-readable version, with client-specific steps, lives on [the agents page](https://trysincerely.com/agents). Client-specific setup pages: [Claude](https://trysincerely.com/integrations/claude-mcp), [Codex](https://trysincerely.com/integrations/codex), [Grok](https://trysincerely.com/integrations/grok), and [Hermes](https://trysincerely.com/integrations/hermes). ## What the agent can do - Build audiences. From your synced [Salesforce](https://trysincerely.com/integrations/salesforce) or [HubSpot](https://trysincerely.com/integrations/hubspot) data, from a [CSV](https://trysincerely.com/integrations/csv), or from a plain description of your ideal customer. From a description, a starter finder run looks for up to 25 US contacts with a work email and an office address, once you confirm what it spends. Your agent can also add prospects it researched itself, 25 at a time. Every sourced address waits for review before it can mail. - Write pieces. Every piece is written for one recipient from CRM context and researched account facts. The agent drafts, you approve. - Prepare campaigns. Formats, budgets, holdout design, launch settings, all staged and waiting for your sign-off. - Check mailability. Addresses are resolved and checked for mailability before anything can print. ## Send one piece from an agent The safe one-off path is five decisions, in order: 1. Ask the agent to find or read the contact. A new or changed address returns to review and postal verification before it becomes mailable. 2. Choose the sender, message strategy, format, and design. A draft design remains a draft until the send is confirmed. 3. Ask the agent to draft the message and show it in full. Edit or reject weak facts before approval. 4. Ask it to prepare the send. Sincerely checks suppression, address status, the rolling frequency cap, blackout dates, price, and workspace authority. 5. Review the confirmation card. It names the recipient, destination, design proof, and current price. Only your click releases the piece. The repeatable Send flow, which is separate from the workspace's guided first send, provides a five-minute cancel window in the mail ledger. Once the vendor has the piece, it cannot be recalled. ## Launch a campaign from an agent A campaign adds an audience, cadence, budget, approved design proofs, and an experiment contract. The agent can prepare those inputs and show the launch preview. Confirmation binds to the audience identity and design proofs it displayed, so changing the recipients or artwork requires a new review. Launch assigns whole accounts to mailed or holdout arms and locks the primary outcome and measurement window. A workspace governance rule can require an additional approval after the first confirmation. The [agent permissions guide](https://trysincerely.com/guides/ai-agent-direct-mail-permissions) explains why the confirmation is structural rather than a prompt instruction. ## What the agent cannot do Spend your money alone. Printing a piece, launching a campaign, changing a plan: each one renders a confirmation card in front of a human. You see exactly what will happen and what it costs before you click. This is not a compliance checkbox bolted on later. It is the design: agents do the work, humans own the money. That gate is why preparation can be broad while release stays narrow. Suppression, postal verification, frequency caps, and budgets still run even after the person confirms. The gate is scoped to what you have to see or pay for. Configuration that spends nothing and mails nothing, such as the brand kit or the workspace name, saves when the agent asks for it. While a card is waiting, the agent can carry on with reads, drafts, designs, and audience work, and `https://trysincerely.com/confirm` lists every card open for you at once. ## Why this beats scripting an API A mail API makes you write and maintain glue code: fetch accounts, format addresses, template copy, handle failures. An MCP agent skips all of it. You say "mail the 20 accounts that went dark this quarter with a piece about their renewal", and the agent queries the CRM, drafts 20 individual pieces, and stages the campaign. You review and approve. The loop that took an afternoon of scripting starts with one sentence and ends with the required human review. If you want deterministic, high-volume, programmatic mail with no human in the loop, an agent is the wrong tool and so is Sincerely. Use a print API like Lob directly. Sincerely is built for targeted B2B mail where a human should see every dollar before it ships. ## Try it without a card New workspaces start on a free audit plan. Your agent can connect, audit mailability, build audiences, and draft pieces before you pay anything. Nothing mails until you choose a [plan](https://trysincerely.com/pricing). Your agent does the work. You sign the check. That is the whole integration. ## Related questions - [Physical mail MCP, how AI agents send real mail](https://trysincerely.com/guides/physical-mail-mcp): What a physical mail MCP server is, what to check before connecting an agent to one, and when wrapping a print API yourself is the better choice. - [How to let an AI agent prepare direct mail safely](https://trysincerely.com/guides/ai-agent-direct-mail-permissions): Let an AI agent research and prepare direct mail, but require a human to confirm recipients and spend before anything prints. - [How to set budgets and spending limits for direct mail](https://trysincerely.com/guides/direct-mail-budget-controls): Set direct mail budgets with term credits, prepaid top-ups, piece caps, approval thresholds, and a final preflight before anything ships. - [Direct mail from a CSV](https://trysincerely.com/integrations/csv): CSV import is a native Sincerely integration. Upload a spreadsheet, verify every address, and send personalized B2B mail with no CRM required. - [Can Claude send a real letter for you?](https://trysincerely.com/guides/can-claude-send-a-letter): Yes. Connect Claude to a physical mail MCP server, review the proof and price, then approve the exact letter before it prints. - [How to let an AI agent spend money safely](https://trysincerely.com/guides/let-an-ai-agent-spend-money-safely): Let an AI agent prepare a purchase, then bind one person's approval to the exact action and enforce every limit on the server. --- Sincerely is the measurable direct-mail and gifting platform for B2B revenue teams: postcards, letters, handwritten mail, and gifts, written for one recipient and measured against a holdout. Contact Sincerely: https://trysincerely.com/contact Agent routing index: https://trysincerely.com/llms.txt