Website · Docs · Dashboard · Knowledge · Changelog · X · LinkedIn
One is the trust layer for AI agents. Connect each app once, in an account that belongs to you, and every agent you run works through that connection: scoped to the action, revocable in one click, and recorded in a complete audit trail. Agents get a grant, never your credentials. One is a passthrough, so your data is never stored.
1,000+ platforms and 130,000+ verified tools, each one tested before an agent runs it.
Point any MCP agent at One's hosted server and sign in when it asks:
$ claude mcp add --transport http one https://mcp.withone.ai/mcpOr install the CLI:
$ npm install -g @withone/cli
$ one init
$ one onboardThe One CLI will prompt you for your API key (get one from the dashboard) and configure One for your agent of choice: Claude Code, Codex, Cursor, Claude Desktop, Windsurf and more. Setup guides for every agent are at withone.ai/mcp.
Use cases shows working agents, each with its own repo, and awesome-one collects what people have built on One.
