All 62 articles, sorted alphabetically
A2A Agent Card Architecture in Depth
A 2500-word walkthrough of A2A agent card architecture: well-known URL, capabilities, skills, endpoints, auth, versioning, signing, registry.
Read article →The A2A Agent Card Specification, Field by Field, in depth: every 1.0 field, what clients do with it, and a card that validates
A field-by-field reference to the A2A 1.0 Agent Card: identity fields, supportedInterfaces and protocol selection, capabilities and extensions, securi…
Read article →A2A artifact exchange architecture
Deep-dive on A2A artifacts: the artifact/part object model, TextPart/FilePart/DataPart types, TaskArtifactUpdate streaming with append and lastChunk, …
Read article →A2A Artifact Types, in depth: text, raw, url and data parts, media types as the type system, multi-part artifacts and output negotiation
A first-principles guide to typing A2A 1.0 artifacts: the four Part content fields, how to choose between them, media types and schema identification,…
Read article →A2A authentication and authorization -- agent identity, delegation, cross-org trust
Deep-dive on A2A auth: agent identity and verifiable credentials, OAuth/OIDC token authentication, scopes and claims, delegation for agents acting on …
Read article →A2A backpressure architecture
Deep-dive on backpressure in Agent2Agent systems: bounded admission windows, credit-based flow control, 429/Retry-After signals, SSE pull cadence and …
Read article →A2A Capabilities, in depth: what each flag gates, server enforcement, extensions, client fallback and capability drift
How A2A 1.0 agent capabilities work as a runtime contract: what streaming, pushNotifications, extendedAgentCard and extensions each gate, the errors a…
Read article →Agent capability architecture, in depth: one capability registry behind the A2A card, routing without a skill id, extensions and enforcement
How to architect what an A2A agent can do on the server side: the three different things called capabilities, an internal capability manifest as the s…
Read article →A2A Capability Discovery, in depth: compiling an Agent Card into an effective per-skill profile, merging the extended card, verifying claims with probes, pinning a capability lockfile and reacting when the card changes
What a client should do after it finds an A2A agent: compile the 1.0 Agent Card into an effective profile per skill (interface, modes, security, capab…
Read article →A2A Capability Negotiation, in depth: choosing binding, version, credentials, media types and update delivery before the first message
How an A2A client and a remote agent agree on how to talk: reading supportedInterfaces to pick a binding and protocol version, the A2A-Version header …
Read article →A2A circuit breaker architecture
Deep-dive on circuit breaking for Agent-to-Agent systems: the per-peer closed/open/half-open state machine driven by a rolling failure rate, failing f…
Read article →Agent conversation architecture, in depth: contexts, multi-turn tasks, history, turn budgets and termination in A2A
How to architect multi-turn conversations between agents with the A2A protocol 1.0: what a conversation is in terms of contextId, tasks and messages, …
Read article →Dead-letter handling architecture
Deep-dive on dead-letter handling for agent-to-agent task delegation: how a task that cannot be delivered or completed after bounded retries is parked…
Read article →A2A Discovery, in depth: well-known cards, registries and direct configuration, a client-side resolver, and fetching untrusted cards safely
How an A2A client finds and prepares to call an agent: the three discovery mechanisms in the specification, a resolver that fetches, validates, verifi…
Read article →A2A Distributed Tracing, in depth: carrying traceparent through A2A requests, streams, long-running tasks and push notifications
How W3C Trace Context travels between A2A agents: the traceparent format and its validation rules, injecting and extracting it on JSON-RPC, gRPC and H…
Read article →A2A error handling -- failures across agent boundaries
Deep-dive on A2A error handling: error types and structured errors (code/message/data), task failure states, retries and idempotency, timeouts, partia…
Read article →A2A Error Recovery, in depth: reconnecting streams, reconciling after a crash, re-delegating failed tasks and compensating side effects
How an A2A 1.0 client recovers after something has gone wrong: resubscribing to dropped streams, a durable delegation journal and reconciler for clien…
Read article →A2A Error Codes, in depth: the nine A2A errors across JSON-RPC, gRPC and HTTP, ErrorInfo reasons, and a client classifier that knows what to retry
A reference and working guide to A2A error codes: the nine A2A-specific errors with their JSON-RPC codes (-32001 to -32009), gRPC and HTTP statuses, t…
Read article →A2A Examples, in depth: six annotated A2A 1.0 wire transcripts, from a one-shot reply to streaming, multi-turn input, push and cancellation
Worked Agent2Agent protocol examples checked against the A2A 1.0 specification: agent card discovery, SendMessage answered by a Message or a Task, an …
Read article →A2A Failure Isolation, in depth: bulkheads per peer, delegation budgets, untrusted remote output, poison-task quarantine and tenant cells
How to stop one failing agent from taking down an agent-to-agent system: a fault taxonomy, per-peer bulkheads with code, deadline, depth and fan-out b…
Read article →A2A File Part Encoding, in depth: inline raw bytes versus URL references, base64 cost, chunked artifacts, signed URL lifetimes and a receiver that fetches safely
How files actually travel in the A2A protocol 1.0: the Part fields raw, url, mediaType and filename; base64 overhead and the proto3 JSON decoding rule…
Read article →A2A idempotency architecture
Deep-dive on idempotency in agent-to-agent systems: client-generated task IDs as the dedup contract, atomic registration and replay, distinguishing re…
Read article →A2A Idempotency, in depth: messageId deduplication, server-assigned task IDs, continuation versus reuse, and safe retries under A2A 1.0
How retries and identifier reuse work in the Agent2Agent protocol 1.0: who mints messageId, taskId and contextId, what the spec does and does not guar…
Read article →A2A input-required State, in depth: when to ask, machine-answerable questions, pausing and resuming on the server, validation, expiry and relaying through orchestrators
How to design and implement the A2A input-required pattern: what the 1.0 specification fixes, when an agent should ask rather than guess or fail, stru…
Read article →A2A JSON-RPC Request/Response Flow, in depth: id correlation, the server dispatch pipeline, a multi-turn wire trace, stream recovery, retries and races
How a request and its responses move through the A2A 1.0 JSON-RPC binding over time: id correlation and notifications, parse-validate-dispatch on the …
Read article →A2A Load Shedding: Admission Control, 429 and Retry-After
Load shedding for agent-to-agent (A2A) systems: why overload turns metastable, priority admission control, bounded queues, fast 429 with Retry-After.
Read article →A2A + MCP Interop Patterns, in depth: four bridge topologies, mapping A2A 1.0 tasks onto MCP tasks, identity across the hop and the failure modes of a bridge
How to connect A2A agents and MCP tools in practice: the four interop topologies, a concept and state mapping between A2A 1.0 and MCP (2025-11-25 and …
Read article →A2A Message Architecture in Depth: the Message object, parts, context and task identifiers, SendMessage and SendStreamingMessage, and the rules that keep multi-turn delegation correct
A first-principles guide to the Agent2Agent (A2A) 1.0 Message: every field, the unified Part with text, raw, url and data content, how contextId and t…
Read article →Inside the A2A Message Envelope, in depth: headers, JSON-RPC frames, request objects, stream and error envelopes
The layers wrapped around an A2A 1.0 message on the wire: HTTP headers and service parameters (A2A-Version, A2A-Extensions), the JSON-RPC 2.0 and HTTP…
Read article →A2A Message Parts, in depth: the wire format across bindings and versions, byte budgets, and a codec that survives version skew
How A2A message parts are encoded on the wire: the protobuf oneof behind every Part, its ProtoJSON form in the JSON-RPC and HTTP+JSON bindings, valida…
Read article →A2A Message Types, in depth: Message, Task, status updates and artifact updates, which content belongs in which object, and a client that reduces the stream
The A2A 1.0 type system as a routing decision: the two responses to SendMessage, the four StreamResponse members, why results belong in Artifacts and …
Read article →A2A Multi-Tenancy: Architecture Deep-Dive
How tenant isolation survives agent delegation chains — signed identity at the edge, scoped token exchange per hop, tenant-keyed state and caches, per…
Read article →Agent negotiation architecture, in depth: offers, counters, policy engines, termination and binding agreements over A2A
How to build agents that negotiate with other organisations' agents over A2A: mapping offer and counter rounds onto one task, a typed offer paylo…
Read article →Agent-to-agent orchestration architecture, in depth: orchestrating remote agents over A2A with tasks, contexts, fan-out, streaming and push
How to build an orchestrator that coordinates remote agents over the A2A protocol: the orchestrator as server and client, mapping a plan onto A2A task…
Read article →A2A Over gRPC, in depth: the A2AService definition, declaring a gRPC interface in the Agent Card, metadata, streaming, rich errors, a working client and server, and running it behind proxies
How the A2A 1.0 gRPC binding works and how to run it: the lf.a2a.v1 A2AService and its eleven RPCs, declaring a GRPC interface in the Agent Card and t…
Read article →A2A Protocol Overview, in depth: the mental model, the objects, the eleven operations and when to use it
A first-principles overview of the Agent2Agent (A2A) 1.0 protocol: what it standardises and what it leaves opaque, how it differs from MCP, the Agent …
Read article →A2A Pagination, in depth: ListTasks cursors, signed page tokens, tasks that move while you page, and keeping long message histories small
How paging works in A2A 1.0: the ListTasks fields, limits and newest-first order, designing an opaque signed keyset page token, why tasks updated mid-…
Read article →Agent-to-Agent Protocol Architecture in Depth: Agent Cards, tasks, streaming, push and the 1.0 wire format
How the A2A 1.0 protocol works end to end: Agent Card discovery, messages and parts, the task state machine, JSON-RPC, gRPC and REST bindings, streami…
Read article →A2A push notification architecture
Deep-dive on A2A push notifications: pushNotificationConfig registration and SSRF-safe URL vetting, the notifier queue and JWT signing, webhook receiv…
Read article →A2A Push Notifications, in depth: configuring webhooks with TaskPushNotificationConfig, the four config operations, credentials, URL policy and receivers
A practical guide to configuring A2A 1.0 push notifications: the TaskPushNotificationConfig and AuthenticationInfo fields, the capability check, regis…
Read article →A2A Rate Limit Response Format, in depth: what the spec leaves open, one rejection shape across JSON-RPC, gRPC and HTTP, and clients that wait the right amount
How an A2A agent should tell a caller it is being rate limited: what A2A 1.0 actually says, separating caller quota from agent overload, a concrete re…
Read article →A2A rate limiting -- protecting agents from overload and abuse
Deep-dive on A2A rate limiting: the overload need, per-agent limits, the token bucket (rate + burst), 429 + Retry-After, distributed limiting (shared …
Read article →Agent registry architecture, in depth: ingesting and verifying A2A Agent Cards, indexing skills, liveness, entitlement-filtered search, tenancy, federation and failure modes
How to build an agent registry for A2A: what the protocol standardises and what it leaves to you, a reference architecture, a data model, an ingest an…
Read article →Agent reputation architecture, in depth: evidence, Bayesian scoring with decay, cold start, rater credibility and Sybil resistance for A2A agents
How to build a reputation layer on top of the A2A protocol, which defines none: what reputation is and is not, the evidence record, verifiable versus …
Read article →A2A Resumable Long-Running Tasks, in depth: what A2A 1.0 guarantees, a durable task store and event log, step checkpoints, leases and fencing, run-once side effects and pauses that last for days
How to make long-running A2A tasks survive agent restarts, deploys and days-long input-required pauses: what the A2A 1.0 specification defines and wha…
Read article →A2A Security, in depth: a threat model for agent-to-agent systems and the controls the specification requires
A threat model for A2A deployments mapped to the specification's security requirements: transport and identity, authorization before lookup, sign…
Read article →A2A Skill Matching, in depth: turning a task into a ranked, calibrated choice of agent skill, or an honest abstention
How an orchestrator matches a task to an A2A agent skill: what AgentSkill fields give you, extracting requirements, hard constraint filtering, hybrid …
Read article →A2A task state -- the lifecycle of a delegated task
Deep-dive on A2A task state: the delegated task as a stateful entity, the state machine (submitted/working/terminal), terminal states (completed/faile…
Read article →A2A State Handoff, in depth: moving responsibility between agents with a versioned envelope, references, redaction and fenced ownership
How to hand work and its state from one A2A agent to another: what A2A 1.0 does and does not carry, designing a versioned handoff envelope as a data p…
Read article →A2A streaming architecture
Deep-dive on A2A's long-running task machinery: message/stream SSE subscriptions, the task state machine, incremental artifact updates, input-req…
Read article →Agent-to-Agent Task Architecture, in depth: the A2A 1.0 Task object, task store, executor, event fan-out and client patterns
How to build the task layer of an A2A agent: the Task object and TASK_STATE values in A2A 1.0, versioned state transitions in a task store, a work que…
Read article →A2A Task Cancellation: tasks/cancel and the Canceled Task State
How A2A task cancellation works: tasks/cancel as a cooperative signal, safe-point checks, saga compensation, sub-agent fan-out, the terminal canceled …
Read article →A2A Task Delegation, in depth: delegation records, the A2A 1.0 task lifecycle, interrupted states, budgets across sub-delegation, and recovery
How one agent hands a unit of work to another over A2A 1.0 and stays responsible for it: the delegation record, SendMessage with returnImmediately, ma…
Read article →A2A Task Lifecycle States, in depth: the nine TaskState values, what the spec fixes versus server policy, how each channel shows transitions, a client reducer, watchdogs and conformance tests
The A2A 1.0 task lifecycle from the caller's side: the nine TaskState values grouped into active, interrupted and terminal; which rules the speci…
Read article →A2A tasks/send vs tasks/sendSubscribe, in depth: blocking, return-immediately and streaming sends in A2A 1.0, reconnecting with SubscribeToTask, and choosing per call
The choice the old tasks/send and tasks/sendSubscribe names described, as it exists in A2A 1.0: blocking SendMessage, SendMessage with returnImmediate…
Read article →A2A timeout handling
Deep-dive on timeout handling in agent-to-agent systems: putting a deadline on every delegated call, propagating a shrinking budget down the chain, wa…
Read article →Agent-to-agent distributed tracing architecture
Deep-dive on distributed tracing for multi-agent systems: root spans, W3C Trace Context propagation across A2A hops, nesting tool and LLM calls as chi…
Read article →A2A Trust Models, in depth: the five trust questions, allowlists versus PKI versus registries versus pinning, and why content trust is a separate problem
How agents decide whom to trust in the Agent2Agent protocol: endpoint, card, caller, delegated authority and content trust, the trust models that answ…
Read article →A2A protocol versioning
Deep-dive on versioning in agent-to-agent systems: advertising protocol versions and capabilities in the Agent Card, negotiating the highest common ve…
Read article →A2A Well-Known Agent Card Discovery, in depth: publishing /.well-known/agent-card.json, one card per origin, caching with ETags, rendering the card at deploy time and checking it like an API
The publisher side of A2A well-known discovery: what the RFC 8615 well-known URI means, why it scopes one card to one origin, serving the Agent Card w…
Read article →Beyond Monoliths: Crafting Interoperable AI Systems with Google’s A2A Agent Architecture
In today’s interconnected digital landscape, building intelligent systems often means orchestrating multiple specialized services.
Read article →Securing A2A Endpoints with OAuth 2.0 and Scoped Permissions
In a burgeoning Agent-to-Agent (A2A) ecosystem, AI agents are constantly discovering and invoking each other's services.
Read article →