Your workspace, inside Claude and ChatGPT.
Docsiv speaks the Model Context Protocol. Add one URL to your AI host and it can search your documents, draft a proposal in your brand voice, and send it for signature, without you leaving the conversation.
https://docsiv.com/api/mcpIn Claude or ChatGPT, add this under Settings, Connectors. You will be asked to sign in and choose a workspace.
Run this in your terminal.
claude mcp add --transport http docsiv https://docsiv.com/api/mcp
Connected in under a minute.
No install, no local server, no config file to hand-edit.
Copy the connector URL
One URL works for everyone, there is nothing per-account to generate. Copy it from the panel above.
Add it as a custom connector
In Claude, go to Settings, Connectors, Add custom connector. In ChatGPT, add it under Connectors in settings. Paste the URL and continue.
Sign in and pick a workspace
You are sent to Docsiv to sign in and approve access. Choose which workspace the host may reach, then approve. That is the whole setup.
Connected? Send this first.
Connecting itself happens in your host’s settings, not in the chat. Once it is authorized, this message gets the assistant oriented in your workspace.
You now have access to my Docsiv workspace over MCP. Start by calling get_workspace_context, then list my clients and ten most recent documents so you have context to work from. Then tell me, in one short list, what you can now do on my behalf: what you can read, what you can create or edit, and what you can send out.
The connector has to be switched on for your workspace first. Open MCP settings
It does the work, not just the talking.
A chat assistant that can only describe a proposal is a writing exercise. This one builds the document, in your workspace, ready to send.
Find anything in the workspace
Search documents, pull a client’s full context, list projects, and read a document’s live content before writing a word.
Create from a sentence
Spin up a proposal, deck, sheet, or invoice from scratch or from one of your templates, using the brand source selected for the document.
Edit in the real editor
Granular edits land in supported native editors, not a copy. Text, decks, spreadsheets, whiteboards, and forms have different editing and export behavior; imported PDFs are not a universal native editor.
Send for signature
Set signing recipients and request approval, then track the state of the request from the same conversation.
Share and publish
Create scoped share links, manage collaborators, and publish a finished document to the web or a client portal.
Export anywhere
Export supported documents in the formats available for their editor, such as PDF, PPTX, DOCX, CSV, PNG, or SVG. The exact format and branding behavior depend on the document type and source.
You approve exactly what it can reach.
Three scopes, shown to you on the approval screen before anything connects.
mcp:readRead
Search and read your documents, clients, projects, and workspace settings.
mcp:writeWrite
Create and edit supported documents on your behalf, inside the workspace you approved.
mcp:shareShare
Create share links and manage who can collaborate on a document.
OAuth 2.1 with PKCE
No API keys to paste or leak. Authorization uses PKCE, refresh tokens rotate on every use, and access tokens are bound to this endpoint.
Scoped to one workspace
You choose the workspace at approval time. The host never sees another workspace, and per-client row-level isolation still applies underneath.
Revoke in one click
Every connected app is listed in your workspace settings. Revoke one and its tokens stop working immediately.
Admins stay in control
Workspace owners turn the connector on, and can narrow which tools are exposed. It is off until someone deliberately enables it.
More on how we protect your work: Security at Docsiv
Questions, answered.
Which AI hosts does this work with?
Any host that supports remote MCP servers over HTTP with OAuth. Claude and ChatGPT both do, and the same URL works for other MCP-compatible clients and IDE agents.
Does it cost Docsiv credits?
That is your call per workspace. In host-token mode the AI host writes the content itself and no Docsiv credits are spent. In credits mode, generation runs Docsiv’s own pipeline and draws on your workspace credits. You pick the mode in workspace settings.
Can it change things without asking me?
Write access only exists if you granted the write scope, and the connector can ask you to confirm before it commits a document or sends an approval request. Everything it does is a normal document action with the usual history behind it.
Is my content used to train models?
No. Content passes through to whichever model your host is running so it can answer you, and Docsiv does not sell or train on your work. See our security page and subprocessor list for the detail.
Still stuck? Talk to us.
Bring your workspace into the chat.
Add one URL and your AI host can search, draft, and send real client documents.