On April 15, 2026, Salesforce announced Headless 360 and declared that everything on its platform is now an API, an MCP tool, or a CLI command that agents can call directly. By Dreamforce in September, the message was blunter: agents are a new class of user, and every SaaS platform has to decide how to expose its value to them. Moor Insights observed that most vendors have answered by shipping a handful of MCP endpoints and calling it done.
For most software, going headless is a plumbing problem. For encryption, it is a trust problem. The UI was where a person decided who receives a file, looked at what was in it, and clicked send. Take the person out and two questions are left: where does the plaintext live, and who holds the key?
Our answer doesn't depend on our portal. Files are encrypted on the device that holds them before anything reaches a SendSafely server, and the decryption key is split so nobody, not even SendSafely, can read the contents on its own. That holds in our REST API, official SDKs, Dropzone widget, and the SendSafely MCP server.
SendSafely has always been API-first
SendSafely has always been an API-first company and an API-first product. We built SendSafely as encryption primitives any client could call, and our own apps were among the first clients. The REST API is on version 2.0. The Client API for Java and .NET is on its third major version, with SDKs for Node.js and Python alongside it, all public on GitHub. Halo and our integrations run on the same primitives. Every one encrypts on the client before a byte reaches our servers, because that is where SendSafely encryption has always happened.
Our MCP server needed no new cryptography or trust model. It exposes operations the API already had: create a package, upload, add a recipient, finalize, list, download.
Here are seven ways to use SendSafely without opening the UI.
1. Let an AI agent send encrypted packages
The SendSafely MCP server lets Claude, Codex, Cursor, Windsurf, VS Code, or any stdio MCP client run a complete encrypted send. Ask it to "send the Q3 audit workpapers to outside counsel, SMS verification, expire in seven days" and it creates the package, uploads the files, adds recipients, sets expiration, watermarking, and download limits, then finalizes.
Encryption happens on your machine. You approve each file upload, and the agent checks with you before it emails anyone the secure link.
2. Have an agent watch a Dropzone like an inbox
In high volume customer service environments, most sensitive data arrives inbound: claim documents, bank statements, signed forms, debug files. The MCP server can list received packages, Dropzone submissions included, and download them. "What came in overnight, and from whom?" becomes a question you ask instead of a portal you check.
Trusted agents can be approved to decrypt received data as well. Reading content is gated twice. The agent needs either the fully keyed secure link or to be registered as a SendSafely Trusted Device, and it asks for approval every time before it downloads or reads a file.
Prefer code to an AI client? "Get Files from Dropzone" in the Code-Examples repo does the same thing.
3. Run Workspaces as data rooms from a terminal
Managing a Workspace as a deal room, vendor exchange or customer project requires configuration and administration. Through the MCP server you can create a Workspace, add a Contracts folder, promote a collaborator to Manager, upload the latest SOW, and pull the folder's activity log in plain language.
4. Encrypt from back end jobs and CI pipelines
Not every headless workflow needs an AI model in it. A nightly job that ships payroll exports to a processor, a pipeline step that delivers build artifacts to a regulated customer, a back-office script that returns statements to clients: these want code, not a conversation.
The official SDKs for Node.js, Python, Java, and .NET handle keycode generation, OpenPGP encryption, request signing, and chunked uploads for large files. Each file is encrypted inside your process before it leaves the host, and a complete send takes a handful of lines.
5. Put encrypted collection inside any front end
Sometimes your product supplies the UI and SendSafely sits behind it. The Dropzone widget is a JavaScript component that turns a web form, customer portal, or onboarding flow into an encrypted upload point. Files are encrypted in the visitor's browser and go straight to SendSafely, so neither your web server nor your form vendor touches the plaintext.
Halo does the same for AI chat. A chatbot can collect a driver's license, bank statement, or HAR file without the file entering the conversation, the model's context, or the ticket.
A Dropzone webhook calls your HTTPS endpoint when a submission is finalized. Before acting on it, re-fetch the package through the signed REST API to confirm it's real.
6. Process files before anyone sees them
SendSafely Actions run automatically when a file arrives, applying your controls before it reaches a ticketing system, an agent's context, or someone's downloads folder.
Our HAR sanitizer proof of concept is one example. An Action hands an uploaded HAR file to an AWS Lambda function in the customer's environment, which strips session cookies, bearer tokens, and SAML assertions and delivers only the sanitized copy. SendSafely never sees the unencrypted content, and no support agent, human or AI, can leak a credential that's already gone.
7. Pipe the audit trail into your SIEM
When agents and scripts do the sending, the audit log is your record. The Audit Log API returns package, user and admin event records with timestamp, source IP, authenticated user, and event details. Send them to your SIEM or export to S3 for indefinite retention.
The MCP server exposes package and Workspace activity too, so you can ask which contracts sent to outside counsel are still unread after 48 hours. Agent sends land in the same trail as portal sends.
Headless by design. Secure by default.
Whether you're building an AI agent, automating a back-end process, or embedding secure file exchange into an existing application, the security requirements don't change. Sensitive files still need to be encrypted before they leave the originating device, access needs to be controlled, and every action needs to be auditable.
That's why SendSafely was built API-first. Our REST API, SDKs, integrations, and MCP server all use the same underlying encryption architecture.
Headless shouldn't mean handing over the keys. With SendSafely, you can remove the UI from the workflow without removing the security controls that protect the data.
The SendSafely MCP server is currently in limited beta and works with any compatible MCP client. Contact your account manager for access and follow the installation guide. Our REST API, SDKs, and code examples are available today.
SendSafely: Integrated File Transfer for the Apps you Love
If you are looking for a secure way to send or receive files with anyone, or simply need a better way to transfer large files, our platform might be right for you.
