You don't have to switch tools to put AI to work on your content. Soontra exposes an MCP (Model Context Protocol) endpoint over streamable HTTP, so the AI client you already use every day — a chat app or a code editor — can connect straight to your workspace and work with what's actually in it.
Once connected, your AI stops guessing about your work and starts answering from it: your projects, your footage, your posts, your numbers.
What connecting an AI client unlocks
A connected client gets roughly 50 tools spanning twelve areas of your workspace: projects, subtasks, comments, files, folders, posts, documents, workflow stages, search, analytics, workspace context, and notifications.
In practice that means you can ask your AI to:
- List your projects, check deadlines, and move work through your workflow stages.
- Search your indexed videos and files for specific moments, then draft a document from what it finds.
- Draft posts and comments that land in Soontra for your review.
- Pull analytics and turn them into a report or content ideas.
Supported clients
Soontra's connection guides cover six clients, split by how they authenticate:
- Sign-in (OAuth) clients: Claude.ai, Claude Desktop, and ChatGPT. You connect by signing in to Soontra — no key to copy. See Connect Claude.ai, Claude Desktop, or ChatGPT.
- API-key clients: Claude Code, Cursor, and VS Code. You create a scoped key and paste one config block. See Connect with API keys.
OAuth vs API keys
OAuth is built for chat apps: the client registers itself, you approve the connection with a Soontra sign-in, and the grant is tied to you and one workspace. API keys are built for developer tools: you create a soontra_sk_ key in Soontra, choose exactly which scopes it carries, and the client sends it as a Bearer token. Keys can be scoped narrowly, given an expiry, and revoked at any time — the right shape for editors, CLIs, and anything you configure by hand.
What connected AIs can and can't do
Connected clients can read what their scopes allow, search your content, and create drafts: projects, documents, post drafts, and comments. What they can never do is take an irreversible action on your behalf.
Every action is also governed by scopes and rate limits, so you decide how much a client can see and change. The full breakdown is in MCP scopes, safety, and rate limits.
Where to set it up
Everything lives in Settings → Connections. That page shows your workspace's MCP server URL, step-by-step guides for each client, and the API-key section where you create and revoke keys. Only workspace owners and admins can connect AI clients.