Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
import { describe, expect, it } from 'vitest'
import {
decodeMcpHostRuntimeJwtClaims,
getJwtIssuedAt,
getJwtRuntimeBinding,
getMcpHostRuntimeCallerKey,
} from './mcpHostRuntimeJwt'
function jwtWithClaims(claims: Record<string, unknown>): string {
const header = Buffer.from(JSON.stringify({ alg: 'none', typ: 'JWT' })).toString('base64url')
const payload = Buffer.from(JSON.stringify(claims)).toString('base64url')
return `${header}.${payload}.sig`
}
describe('mcpHostRuntimeJwt', () => {
it('uses hostRefs[0] as the canonical caller and runtime binding over sub', () => {
const token = jwtWithClaims({
sub: 'mcp-host/chatllm',
hostRefs: ['chatllm', 'sandbox-recipes/secondary'],
recipeNamespace: 'mcp-host',
recipeName: 'standalone',
exp: 123,
iat: 99,
})
expect(decodeMcpHostRuntimeJwtClaims(token)).toEqual({
callerKey: 'chatllm',
hostRef: 'chatllm',
recipeNamespace: 'mcp-host',
recipeName: 'standalone',
exp: 123,
iat: 99,
})
expect(getMcpHostRuntimeCallerKey(token)).toBe('chatllm')
expect(getJwtIssuedAt(token)).toBe(99)
expect(getJwtRuntimeBinding(token)).toEqual({
hostRef: 'chatllm',
recipeNamespace: 'mcp-host',
recipeName: 'standalone',
})
})
it('returns null caller key without falling back to sub or secondary hostRefs', () => {
const token = jwtWithClaims({
sub: 'mcp-host/chatllm',
hostRefs: [' ', 'sandbox-recipes/secondary'],
recipeNamespace: 'mcp-host',
recipeName: 'standalone',
})
expect(getMcpHostRuntimeCallerKey(token)).toBeNull()
expect(getJwtRuntimeBinding(token)).toBeNull()
})
it('returns null claims and no runtime binding for malformed tokens', () => {
expect(decodeMcpHostRuntimeJwtClaims('not-a-jwt')).toEqual({
callerKey: null,
hostRef: null,
recipeNamespace: null,
recipeName: null,
exp: null,
iat: null,
})
expect(getMcpHostRuntimeCallerKey('not-a-jwt')).toBeNull()
expect(getJwtRuntimeBinding('not-a-jwt')).toBeNull()
})
})