Use Ceven from your AI assistant
Ceven runs a hosted MCP server. Add it to Claude, Codex, or another OAuth-capable MCP client with just a link, authorize once in your browser, and create agents, use workspace-connected apps, build automations, and inspect runs — all from chat. No API key to paste.
How it works
You add the Ceven MCP server to your client using only its URL. The first time the client connects, it sends you to a Ceven consent screen in your browser. You sign in with your normal Ceven account and click Authorize. That’s it — the client receives a short-lived token that it refreshes automatically. You never paste a password or a long-lived key.
Under the hood this is standard OAuth 2.1 with PKCE. The token is scoped to your account, refreshes on its own, and is revocable at any time from your MCP client.
Your client permanently receives only four small tools: ceven_search, ceven_describe, ceven_read, and ceven_act. Ceven searches its permission-aware operation registry first and loads one exact schema only when needed, so adding new agents, apps, and automation operations does not inflate every conversation.
https://api.ceven.io/api/v1/mcpClaude Code
1. Add the server — run this in your terminal:
claude mcp add --transport http --scope user ceven.io https://api.ceven.io/api/v1/mcp2. Start an interactive session:
claude3. Inside the session, run
/mcp, select ceven.io, and choose Authenticate.4. A browser tab opens to Ceven — click Authorize. The token is managed and refreshed for you from then on.
Codex
1. Add the hosted Streamable HTTP server:
codex mcp add ceven --url https://api.ceven.io/api/v1/mcp2. Start OAuth authorization:
codex mcp login ceven3. Your browser opens the Ceven consent screen. Sign in, select the workspace, and click Authorize.
4. Confirm the connection:
codex mcp list
Codex desktop, CLI, and the IDE extension share this MCP configuration and its OAuth connection.
Claude Desktop & claude.ai
1. Open Settings → Connectors → Add custom connector.
2. Paste this as the server URL:
https://api.ceven.io/api/v1/mcp3. Save, then click Connect — a browser window opens to Ceven; click Authorize.
ChatGPT
1. Open Settings → Connectors (or a custom GPT → Actions) → Add.
2. Paste this as the MCP server URL:
https://api.ceven.io/api/v1/mcp3. Connect — a browser window opens to Ceven; click Authorize.
MCP connector support varies by ChatGPT plan and surface. If you don’t see custom connectors, use Claude, Claude Code, or Codex, which support the flow above today.
What you can do
Once connected, your assistant can use the Ceven tools on your behalf. The connector acts as you — it can do what you can do in the Ceven app.
Agents
- Create a private, teammate-shared, or organization agent from plain-language instructions
- Bind exact workspace app accounts and Libraries, then publish the runtime
- Discover and run agents you have an explicit use grant for
Connected apps
- See which apps you’ve connected and what actions are available
- Search and describe one provider action only when it is needed
- Run an action against the exact account selected; app provisioning remains admin-only
Automations
- Create or update an automation from a high-level goal
- Schedule, pause, resume, or run an automation
- Use the canonical Workflow Builder compiler instead of manufacturing an internal graph
Runs & observability
- List runs and inspect their step-by-step trace and output
- Approve a waiting run or cancel one you are allowed to control
- Use idempotent writes so client retries do not duplicate work
Security & revoking access
Connecting never asks for a password or a long-lived API key. The client gets a short-lived access token plus a refresh token, both tied to your account and both revocable.
You can disconnect Ceven from your MCP client at any time. Revoking takes effect immediately — the client has to re-authorize in the browser to reconnect.
FAQ
Do I need to paste an API key or token?
No. You add only the server URL and authorize in your browser. Ceven manages a short-lived, auto-refreshing token for the client.
What is MCP?
The Model Context Protocol is an open standard that lets AI assistants connect to external tools and data. Ceven runs a hosted MCP server so your assistant can use Ceven directly.
Is my data safe?
The connector acts as your account with a revocable, short-lived token — the same permissions you already have in the Ceven app. You can disconnect the client at any time.
How do I disconnect a client?
Remove the Ceven MCP server from your client. Its token stops working immediately and it must re-authorize to reconnect.
Which clients are supported?
Any MCP client that supports remote (HTTP) servers with OAuth. Claude Code, Claude (Desktop & claude.ai), and Codex work today; ChatGPT support depends on your plan and surface.