Connect AI clients with scoped API tokens.
ServerChirp exposes its automation catalog to compatible AI clients through Model Context Protocol (MCP). The endpoint is stateless Streamable HTTP at https://serverchirp.com/api/mcp and accepts Bearer API tokens only.
MCP access requires Business or higher, where API access is enabled, and a token containing the scopes for the operations the client may use. The tool list is filtered by those scopes, so an unavailable operation is not advertised to the model.
Create a scoped API token in ServerChirp, then configure the client to send it as the Authorization Bearer token. For clients that support environment interpolation, keep the token outside the checked-in configuration.
{
"mcpServers": {
"serverchirp": {
"type": "http",
"url": "https://serverchirp.com/api/mcp",
"headers": {
"Authorization": "Bearer ${SERVERCHIRP_TOKEN}"
}
}
}
}Client configuration formats differ. Use its documented secret or environment mechanism rather than putting the token directly in a shared file.
MCP tools project from the same catalog used by the ServerChirp assistant. Each tool declares the exact API scope used by the /api/v1 endpoint that performs the operation.
API tokens are team-wide. Resource-specific member grants do not apply to a token, so scope the token narrowly and create a separate token for each client or automation boundary.
A :write scope allows the client to change infrastructure. Whether an external client asks for confirmation is controlled by that client. Grant read scopes first, inspect the available catalog, and add individual write scopes only when the workflow needs them.
Revoke the token from ServerChirp to disconnect the client immediately. Rotate a token if it appears in logs, shell history, a repository, or a configuration shared with someone outside the team.