Create, send and store contracts with your agents
Updated
Short answer
Connect your contract tool to your AI agent through MCP and the whole contract job moves into the chat: you ask for a contract in plain words, the agent drafts it from your template, you edit it inline and press Send, and the signed copy is stored where you can ask about it later. It works in Claude, ChatGPT, Codex and Cursor.
Why contracts, and why now
Agencies and consultancies already spend their day in an AI assistant: writing proposals in Claude, planning in ChatGPT, building in Codex or Cursor. Contracts are the odd one out. The client says yes in a chat or an email, and then someone opens a separate tool, finds the template, retypes the client's details and the price, exports a PDF and uploads it for signing.
MCP, the Model Context Protocol, removes that detour. Your agent can use your contract tool directly, so the contract is made where the yes was.
What it looks like
| You say | What happens |
|---|---|
| "Make a consulting agreement for Fjord Studio from our standard template, 120 hours from 1 December." | A draft is created from your template, with the client and fields filled in. Nothing is sent. |
| "Make the notice period 60 days." | The agent changes the draft. You see the change in the contract. |
| "Send it to their CEO, Jane Smith." | The contract is sent for signature and locked. |
| "Which contracts are still waiting for a signature?" | A list of sent, unsigned contracts and who they're waiting on. |
| "Which contracts end in the next three months?" | Signed contracts with end dates coming up. |
| "Import this signed PDF for Northwind." | An old contract is stored next to the new ones. |
In Claude, ChatGPT and Cursor the contract appears in the chat as a document: edit the text, click a field to fill it in, select a sentence to ask for a change, and press Send. In Codex the agent shows it as text, with a link to the draft.
The three rules that make it safe
- Drafts come from your templates. The legal text is yours. The agent fills in the client, scope, price and dates. See why your agent should draft from your templates.
- You decide when it's sent. The agent never sends on its own. Sending happens when you press Send or tell it to.
- It acts as you. Your role, your workspace, read-only if you prefer, no delete tools, and every change logged. See is it safe to let an AI agent send contracts?
Getting started
- Claude: add Pledge under Settings, then Connectors. Step by step: how to create and send a contract with Claude.
- ChatGPT: add the Pledge app. See draft, edit and send contracts in ChatGPT.
- Codex and Cursor: add the Pledge MCP server or plugin. See send a statement of work from Codex or Cursor.
The full set-up for each app is in Pledge's agent guide. Then try a few of the 15 contract prompts for agencies.
Beyond contracts
Once a contract is signed, the same agent can do more if you turn on Pledge's other modules: log time against the project, check who has room next week, or turn a new lead into an offer. Contracts come first. The rest is there when you need it.
Questions
- Which AI agents work with Pledge?
- Claude, ChatGPT, Codex and Cursor, and other assistants that support MCP servers.
- Can the agent send a contract without me knowing?
- No. Drafts can be edited freely, but a contract is only sent when you press Send or ask the agent to send it. Sent contracts are locked.
- Do I need to be technical to set it up?
- No. You add Pledge as a connector or app in your assistant, sign in, and approve access. There is no code involved.
- Is my data used to train the AI model?
- That depends on your AI provider and your plan with them. Pledge doesn't receive your conversations, only the actions the agent takes. Check the data settings in your assistant.
Try Pledge
Contracts, time, plans and leads for small agencies, with an MCP server for Claude and ChatGPT. Pledge is in private beta: leave your email and we will send an invite.