Connect any remote MCP client to Primeta with a long-lived Primeta API token. Use this when OAuth isn't available or you want a fully headless setup (CI, scripts, shared workspace configs).
When to use this
- Your client supports remote MCP servers but doesn't do browser OAuth
- You want a static Bearer-token config you can commit (carefully) or sync to multiple machines
- You're wiring Primeta into a coding agent like Claude Code via its project-level MCP config
Setup
- Generate your Primeta API token at connection setup → API Token tab.
- Add Primeta as a remote MCP server with a Bearer header. No npm install — it's the same
/mcpendpoint as OAuth MCP, just with a static token in place of the browser login:json { "mcpServers": { "primeta": { "type": "http", "url": "https://primeta.ai/mcp", "headers": { "Authorization": "Bearer YOUR_PRIMETA_API_TOKEN" } } } } - Drop that into your client's MCP config (for Claude Code, that's
.mcp.jsonat the project root — add it to.gitignore; it holds your token). Start the client. Primeta's tools become available.
The connection setup page pre-fills this snippet with your real token under the API Token tab.
How it works
Your client hits Primeta's remote MCP endpoint over HTTP on every tool call, authenticated by the Bearer token. Each call creates or reuses an MCP session under your user account — so persona switches, sent messages, and animation triggers all land on your Primeta dashboard in real time.
A channel is different: that's the persistent WebSocket connection a local coding agent (Claude Code, OpenClaw) opens. Token MCP and OAuth MCP create MCP sessions, not channels — but both kinds of connection show up as cards on your dashboard.
See conversation sessions for how sessions behave once connected.