Use multiple JWT Bearer Authentication

You can totally achieve what you want:

services
    .AddAuthentication()
    .AddJwtBearer("Firebase", options =>
    {
        options.Authority = "https://securetoken.google.com/my-firebase-project"
        options.TokenValidationParameters = new TokenValidationParameters
        {
            ValidateIssuer = true,
            ValidIssuer = "my-firebase-project"
            ValidateAudience = true,
            ValidAudience = "my-firebase-project"
            ValidateLifetime = true
        };
    })
    .AddJwtBearer("Custom", options =>
    {
        // Configuration for your custom
        // JWT tokens here
    });

services
    .AddAuthorization(options =>
    {
        options.DefaultPolicy = new AuthorizationPolicyBuilder()
            .RequireAuthenticatedUser()
            .AddAuthenticationSchemes("Firebase", "Custom")
            .Build();
    });

Let's go through the differences between your code and that one.

AddAuthentication has no parameter

If you set a default authentication scheme, then on every single request the authentication middleware will try to run the authentication handler associated with the default authentication scheme. Since we now have two possible authentication schemes, there's no point in running one of them.

Use another overload of AddJwtBearer

Every single AddXXX method to add an authentication has several overloads:

• One where the default authentication scheme associated with the authentication method is used, as you can see here for cookies authentication
• One where you pass, in addition to the configuration of the options, the name of the authentication scheme, as on this overload

Now, because you use the same authentication method twice but authentication schemes must be unique, you need to use the second overload.

Update the default policy

Since the requests won't be authenticated automatically anymore, putting [Authorize] attributes on some actions will result in the requests being rejected and an HTTP 401 will be issued.

Since that's not what we want because we want to give the authentication handlers a chance to authenticate the request, we change the default policy of the authorization system by indicating both the Firebase and Custom authentication schemes should be tried to authenticate the request.

That doesn't prevent you from being more restrictive on some actions; the [Authorize] attribute has an AuthenticationSchemes property that allows you to override which authentication schemes are valid.

If you have more complex scenarios, you can make use of policy-based authorization. I find the official documentation is great.

Let's imagine some actions are only available to JWT tokens issued by Firebase and must have a claim with a specific value; you could do it this way:

// Authentication code omitted for brevity

services
    .AddAuthorization(options =>
    {
        options.DefaultPolicy = new AuthorizationPolicyBuilder()
            .RequireAuthenticatedUser()
            .AddAuthenticationSchemes("Firebase", "Custom")
            .Build();

        options.AddPolicy("FirebaseAdministrators", new AuthorizationPolicyBuilder()
            .RequireAuthenticatedUser()
            .AddAuthenticationSchemes("Firebase")
            .RequireClaim("role", "admin")
            .Build());
    });

You could then use [Authorize(Policy = "FirebaseAdministrators")] on some actions.

A final point to note: If you are catching AuthenticationFailed events and using anything but the first AddJwtBearer policy, you may see IDX10501: Signature validation failed. Unable to match key... This is caused by the system checking each AddJwtBearer in turn until it gets a match. The error can usually be ignored.

This is an extension of Mickaël Derriey's answer.

Our app has a custom authorization requirement that we resolve from an internal source. We were using Auth0 but are switching to Microsoft Account authentication using OpenID. Here is the slightly edited code from our ASP.Net Core 2.1 Startup. For future readers, this works as of this writing for the versions specified. The caller uses the id_token from OpenID on incoming requests passed as a Bearer token. Hope it helps someone else trying to do an identity authority conversion as much as this question and answer helped me.

const string Auth0 = nameof(Auth0);
const string MsaOpenId = nameof(MsaOpenId);

string domain = "https://myAuth0App.auth0.com/";
services.AddAuthentication()
        .AddJwtBearer(Auth0, options =>
            {
                options.Authority = domain;
                options.Audience = "https://myAuth0Audience.com";
            })
        .AddJwtBearer(MsaOpenId, options =>
            {
                options.TokenValidationParameters = new Microsoft.IdentityModel.Tokens.TokenValidationParameters
                {
                    ValidateAudience = true,
                    ValidAudience = "00000000-0000-0000-0000-000000000000",

                    ValidateIssuer = true,
                    ValidIssuer = "https://login.microsoftonline.com/9188040d-6c67-4c5b-b112-36a304b66dad/v2.0",

                    ValidateIssuerSigningKey = true,
                    RequireExpirationTime = true,
                    ValidateLifetime = true,
                    RequireSignedTokens = true,
                    ClockSkew = TimeSpan.FromMinutes(10),
                };
                options.MetadataAddress = "https://login.microsoftonline.com/9188040d-6c67-4c5b-b112-36a304b66dad/v2.0/.well-known/openid-configuration";
            }
        );

services.AddAuthorization(options =>
{
    options.DefaultPolicy = new AuthorizationPolicyBuilder()
        .RequireAuthenticatedUser()
        .AddAuthenticationSchemes( Auth0, MsaOpenId )
        .Build();

    var approvedPolicyBuilder =  new AuthorizationPolicyBuilder()
           .RequireAuthenticatedUser()
           .AddAuthenticationSchemes(Auth0, MsaOpenId)
           ;

    approvedPolicyBuilder.Requirements.Add(new HasApprovedRequirement(domain));

    options.AddPolicy("approved", approvedPolicyBuilder.Build());
});