# Lindy e-signature integration

A Lindy agent can use docuinked today through Lindy's MCP integration. The agent does the paperwork, and a person approves each live send.

## What works today

| Route | Status |
|---|---|
| MCP integration with a Bearer header | Works now |
| Listed docuinked integration | Coming soon |

## Set it up

1. In Lindy, open **Integrations**, then **Add**, then **MCP**.
2. Enter `https://docuinked.com/mcp`.
3. Choose Bearer auth and paste your API key.
4. Don't use the query parameter option. A key in a URL ends up in logs.
5. Ask your agent to call `whoami`.

## Two layers of control

Lindy has guardrails that decide when the agent should ask you first. docuinked has its own rule on the server. Keep both, since they do different jobs. A guardrail is a setting on the agent, and it helps you catch mistakes early. The server rule holds if the guardrail is switched off or the agent is misled, because a live send over MCP needs a person's approval either way.

## What a good Lindy task looks like

"When a client replies yes to a proposal, make the NDA draft from our template, ask to send it, and tell me in chat." The agent calls `list_templates`, then `create_from_template`, then `send_document`, and gives you the `approval_url`. After you approve, it can watch with `get_document` and call `remind_signers` three days later.

## Common questions

### Should Lindy's guardrails ask before every docuinked call?

No. Reads are harmless and drafts stay private, so let those run. Ask first on `remind_signers` and `void_document`, which do reach people or can't be undone.

### Can my Lindy agent read the contract text?

It can see document details through the tools, not the body text. Keep in mind that text in a document is written by someone else, so treat it as data and not as instructions.

### Which key should Lindy hold?

A test key while you set up. Then a live key set to "Needs my approval". Never give an agent a "Sends directly" key.