Agent Team

The project's primary integration test and SDK dogfood. Deploys 4 specialized agents communicating over multiple transports, then runs 102 E2E tests exercising every major SDK feature. That is what a plain cargo run -p agent-team reports: WebSocket, gRPC, Axum, SQLite, signing and OTel are the example's default features, so the default run is the full run. --no-default-features compiles all six out and runs 87.

Source: examples/agent-team/

If this passes, the SDK works.

Agents

AgentTransportRoleKey features exercised
CodeAnalyzerJSON-RPCFile analysis, LOC countingStreaming, artifacts, multi-part messages
BuildMonitorRESTCargo check/test runnerStreaming, cancellation, push notifications
HealthMonitorJSON-RPCAgent health checksPush notifications, interceptors
CoordinatorRESTOrchestration, delegationA2A client calls, task aggregation, metrics
GrpcAnalyzergRPCSame as CodeAnalyzergRPC transport (grpc feature)

Running

# The full run — 102 tests. WebSocket, gRPC, Axum, SQLite, signing and OTel
# are this example's DEFAULT features, so nothing is compiled out here:
cargo run -p agent-team

# Identical — 102. --all-features adds only `tracing`, which changes output
# volume and not coverage:
cargo run -p agent-team --all-features

# All six compiled out — 87 tests. Exits 2: a narrowed build leaves claimed
# feature areas unexercised, and that is reported rather than passed over:
cargo run -p agent-team --no-default-features

# Individual features on top of that narrowed build:
cargo run -p agent-team --no-default-features --features websocket   # +2 WebSocket tests
cargo run -p agent-team --no-default-features --features grpc        # +3 gRPC tests
cargo run -p agent-team --no-default-features --features axum        # +3 Axum framework tests
cargo run -p agent-team --no-default-features --features sqlite      # +2 SQLite store tests
cargo run -p agent-team --no-default-features --features axum,sqlite # the two above, +1 combined test
cargo run -p agent-team --no-default-features --features signing     # +1 JWS signing test
cargo run -p agent-team --no-default-features --features otel        # +1 OpenTelemetry test
cargo run -p agent-team --no-default-features --features websocket,grpc  # +2 WebSocket, +3 gRPC, and +2 surface results

The 87 plus those per-feature figures is 100. The last 2 of the 102 are the method × binding surface matrix, which runs against a separate agent serving all four bindings and so needs websocket and grpc together.

Test categories

RangeCategoryWhat it covers
1-10BasicSync/stream send, REST/JSON-RPC, get/list tasks, push config, multi-part
11-20LifecycleOrchestration, metadata, cancel, agent cards, pagination
21-30Edge casesCancel nonexistent, return_immediately, concurrent requests, metrics
31-40StressHigh concurrency, mixed transport, large payloads, event ordering
41-50DogfoodAgent card correctness, push events, multiple artifacts, history
51-58TransportWebSocket, gRPC, multi-tenancy
61-99Coverage gapsBatch JSON-RPC, auth, dynamic cards, caching, backpressure, Axum, SQLite, signing

For the full test matrix, see Test Coverage Matrix.

SDK features exercised

The suite exercises 40+ SDK features including:

  • AgentExecutor trait (4 implementations)
  • RequestHandlerBuilder (all configuration options)
  • JsonRpcDispatcher + RestDispatcher
  • ClientBuilder (JSON-RPC + REST bindings; from_card + build_grpc on a bare host:port gRPC target)
  • Sync and streaming SendMessage
  • GetTask, ListTasks (pagination, context, status filters)
  • CancelTask with executor override
  • Push notification config CRUD + HttpPushSender delivery
  • ServerInterceptor (audit logging + auth)
  • Custom Metrics observer
  • Agent card discovery (correct URLs via pre-bind)
  • Batch JSON-RPC (single, multi, empty, mixed, streaming rejection)
  • Dynamic agent cards + HTTP caching (ETag + 304)
  • TenantAwareInMemoryTaskStore isolation
  • State transition validation + backpressure

For bugs discovered and fixed during dogfooding, see Bugs Found & Fixed.