AI agents can qualify a lead, finish an onboarding conversation, or resolve a support ticket entirely on their own. Then they hit a wall: the task actually needed a real document — a proposal, an offer letter, a certificate, an invoice — and the agent can only describe what that document should say. Someone still has to open a template, fill it in by hand, and send it.
ActiveMerge’s MCP server closes that gap. It gives any MCP-compatible AI agent — Claude, ChatGPT, or another client — direct access to the real templates and document-generation pipeline already in your ActiveMerge account, with a human approval step built into every single document before it’s generated.
What MCP is, and why it matters here
MCP (Model Context Protocol) is an open standard that lets AI assistants connect to outside tools and data sources in a structured way, instead of every app needing a custom one-off integration built for it.
ActiveMerge runs its own MCP server, so an agent doesn’t have to guess at your document format or have you paste data into a chat — it can see your actual saved templates, know exactly what fields each one needs, and produce the real file.
How it works
- Connect. Add ActiveMerge as a remote MCP server in your client (Claude, ChatGPT, or another) using this Streamable HTTP endpoint:
https://app.activemerge.com/mcpChoose Authenticate, sign in with your ActiveMerge account, and approve access. No API key, no client ID — the connection uses whatever plan and credits that account already has.
- The agent reads your templates. Once connected, it can list the templates in your account and see the fields each one requires.
- The agent prepares the exact request. Using data it collected during the conversation or a workflow, it puts together the specific document request — which template, which values, which output format.
- You approve. ActiveMerge returns an approval link. Nothing is generated and no credits are used until that exact request is opened and confirmed.
- It generates. Once approved, the document is created, and the agent can check the job status and hand back the download link.
Why use it
- The agent’s output becomes a finished, on-brand document — not a draft of text you still have to turn into a real file yourself.
- Nothing ships without a human confirming it first, so an agent can work autonomously through the rest of a task without the risk of the wrong document reaching a client.
- No developer time spent building a custom integration between your AI agent and your document system — the connection uses a standard already built into the client.
- The agent works with what you already built. It reads your existing templates and account instead of needing to be taught your document formats from scratch.
- The task actually finishes in one place. Whatever the agent was already doing — qualifying a lead, wrapping up onboarding, closing a support ticket — ends with a real deliverable, instead of a manual handoff to whoever normally builds that document.
Use cases
Sales: qualify a lead, then prepare the proposal
Once an agent has qualified a lead and collected the account details, it can inspect your Sales Proposal template and prepare a proposal from that data.
“When a sales lead is qualified, prepare a proposal from the Sales Proposal template using the account details we collected.”
Onboarding: finish setup, then generate the paperwork
An onboarding agent that’s gathered a new customer’s or employee’s details can prepare the agreement, welcome letter, certificate, or account document that normally follows.
“Create an employment offer letter from the Offer Letter template for the candidate we just approved, starting the 1st.”
Support & operations: resolve the request, then produce the document
A support or ops agent can turn the outcome of its work into a confirmation, summary, invoice, or report, without a second tool or a manual step afterward.
“Create a PDF invoice from my Invoice template for Acme Ltd. for $500.”
Turn a brief into a brand-new template
Describe the layout, content, placeholders, and format, and ActiveMerge drafts a new template for review.
“Create a DOCX project proposal template with client details, scope, timeline, and pricing.”
Creating the AI draft costs 5 credits. You review and approve it at the ActiveMerge link before it’s saved — your existing templates are never touched in the process.
What your agent can do once connected
- List your templates and read the fields each one requires
- Prepare a document from a template and collected data, returning an approval link
- Generate the document once you approve the exact request
- Check a job and retrieve the download link when it’s ready
- Draft a new template in DOCX or PPTX for you to review and save
A few prompts worth trying once you’re connected:
- “What fields does my Invoice template require?”
- “Create a PDF invoice from my Invoice template for Acme Ltd. for $500.”
- “Create a course completion certificate from my Certificate template for Jordan Lee, dated today.”
- “Check the status of my generated proposal.”
Staying in control
- OAuth, not API keys. The connection is tied to the ActiveMerge account you signed in with, and you can revoke it anytime from your account.
- Every generation needs approval. The agent can prepare a request, but nothing generates — and no credits are used — until the exact request is confirmed.
- Every new template needs approval. An AI-drafted template is only saved once you review and approve it; existing templates are never silently edited.
- Zapier is unaffected. If you already use the Zapier integration, it keeps running on its own API key — MCP is a separate connection.
Getting started
Connecting takes about the same five minutes as generating your first document by hand. Start at the MCP overview for the connection details and endpoint, or see MCP use cases for more example prompts.
Frequently asked questions
Do I need an API key to connect an AI agent to ActiveMerge?
No. The MCP connection uses OAuth — you sign in with your ActiveMerge account and approve access, the same way you’d authorize any other app.
Can an AI agent generate and send a document without anyone checking it?
No. Every document generation requires you to open an approval link and confirm the exact request before it’s created and before any credits are used.
Will connecting MCP break my existing Zapier integration?
No. Zapier continues to run on its own API key. MCP is a separate connection layered on top, not a replacement.
Can an agent edit my existing templates?
No. Agents can draft brand-new templates for your review, but existing templates are never edited automatically — a new AI-drafted template only gets saved after you approve it.
Which AI clients does this work with?
Any MCP-compatible client that supports custom connectors over Streamable HTTP — Claude and ChatGPT both support this today, on plans that include custom MCP connectors.
