Connecting your client
Wire Claude (or any MCP client) into your agent's MCP server in a couple of minutes.
You'll need a Agent API Token (okm_…) and the endpoint URL:
https://mcp.onkey.chat/functions/v1/mcp-server
Claude Desktop / Claude Code
Add the server to your MCP client config. For Claude Desktop, edit claude_desktop_config.json; for Claude Code, edit your claude.json (project or user level). The entry is the same in both:
{
"mcpServers": {
"onkey": {
"type": "http",
"url": "https://mcp.onkey.chat/functions/v1/mcp-server",
"headers": {
"Authorization": "Bearer okm_<your-token-here>"
}
}
}
}Replace okm_<your-token-here> with the token OnKey staff issued you, then restart the client.
Confirm the connection
Ask the assistant "what MCP tools do you have?" (or run tools/list). You should see the tools your token's scope grants — a read token shows the five read tools; a full token shows all of them.
Other MCP clients
Any client that supports MCP over Streamable-HTTP connects the same way: point it at the endpoint URL and set an Authorization: Bearer okm_… header.
Token security
- The token is a bearer credential — anyone who has it can act as you. Keep it out of version control, client-side code, and shared documents.
- Tokens are per-agent and scoped: a token for one agent can't touch another company's agent.
- Every action is attributed to the Admin the token was issued for — the same attribution as emailing your Virtual Operations Agent.
- If a token is exposed, contact OnKey staff to revoke it. Revocation is immediate.
Related
Updated about 1 hour ago

