CTO · VP Engineering · A2A Agent Protocol

Agents that stop at
your org boundary.
Operations that don't.

Most enterprise AI agents are wired together by your engineering team — custom integrations, hardcoded endpoints, no discovery. When a new specialist agent joins the fleet, someone builds the connection. Scaling the fleet means scaling the integration work.
A2A (Agent-to-Agent) is the open JSON-RPC 2.0 protocol that changes this. Every compliant agent publishes a machine-readable Agent Card. The orchestrator discovers capabilities, delegates tasks, and receives durable status updates — without out-of-band configuration or proprietary SDKs.
PLRX runs every specialist agent as a fully compliant A2A server — in production, against real operational workflows. The protocol is not a roadmap item. It is the coordination layer behind every mission running today.

94% autonomous resolution From $0.99 per mission Enterprise AgenticA2A · MCP · Open protocols

Book a Scoping Call

Discuss the architecture.

Tell us which operational workflow you are evaluating for A2A-based agent deployment. Proof of concept in 2–3 weeks — production in 12 weeks.

First Name Required.

Last Name Required.

Work Email Please enter your corporate email address.

Company Required.

Job Title Required.

Something went wrong. Please try again or email [email protected].

By submitting you agree to our Privacy Policy. We never sell your data.

Request received.

Someone from the PLRX team will be in touch within two business days to schedule your scoping call.

Agents Without A2A vs PLRX with A2A

Without A2A

How agents coordinate

API calls wired by your engineering team. Each agent integration is custom-built. Adding a specialist agent means building a new integration. Nothing is discoverable — the orchestrator must be told about every agent it coordinates with.

What happens at the boundary

Agents stop at your organisation's edge. Coordinating with a payer system, a custodian platform, or a supplier portal requires a human to cross the boundary — or a brittle scraper that breaks when the UI changes.

Long-running task handling

Session-based. The orchestrator submits a task and waits. If the response takes days, the session either polls continuously (resource drain) or times out and loses the task state.

Fleet visibility and governance

Per-agent logging. Each agent in the fleet generates its own logs. Reconstructing what the fleet did on a specific mission requires aggregating records from multiple sources.

Custom integrations · No discovery · Organisational boundary requires human handoff · Per-agent logs

PLRX with A2A

How agents coordinate

Agent discovery via Agent Card. Every A2A agent publishes a machine-readable manifest at /.well-known/agent.json — name, endpoint, declared skills, authentication. The orchestrator discovers capabilities before contact. No out-of-band configuration.

What happens at the boundary

Agents cross boundaries natively. A PLRX agent coordinating with a payer system, a broker-dealer platform, or a specialist sub-agent uses the same A2A protocol on both sides. The boundary is the protocol. Not a human handoff.

Long-running task handling

Durable task lifecycle. A task submitted via message/send moves through submitted → working → input_required → completed. Each transition is persisted by the Durable State Machine. The task suspends in input_required for days without losing context. tasks/resubscribe re-attaches to the running task.

Fleet visibility and governance

Unified mission audit trail. Every A2A task delegation — orchestrator to specialist, specialist result back to orchestrator — is logged in the PLRX WORM audit trail. The complete fleet activity for any mission is queryable from a single record.

Agent Cards · Open discovery · Boundary-crossing native · Unified mission audit trail · Durable task lifecycle

A2A In Production — Live Sequence from a PLRX Mission

The sequence below is taken directly from a live PLRX production mission. Step 1 discovers the specialist agent. Step 2 delegates the task. Step 3 shows the task entering a durable wait state — suspended without resource consumption — and resuming three days later when the clinical document arrives.

# Step 1 — Discover agent capabilities

GET https://agent.plrx.ai/.well-known/agent.json

# Response — agent manifest

{

"name": "clinical-documentation-specialist-agent"

"skills": ["document_collection", "document_follow_up", "get_status"]

"authentication": { "schemes": ["bearer"] }

}

── Step 2 — Submit task via message/send ─────────────────────────

POST / { "method": "message/send" }

"skill": "document_collection"

"order_id": "ORD-787"

"idempotency_key": "<deterministic>"

── Step 3 — Task enters durable wait state ───────────────────────

POST / { "method": "tasks/get", "id": "task-doc-787" }

{ "status": { "state": "input_required" }

"message": "Awaiting F2F document from provider" }

# Task suspended durably. No resources consumed. Resumes when F2F arrives.

◎ task-doc-787 state: input_required · day 0

◎ task-doc-787 state: input_required · day 3 (F2F received)

✓ task-doc-787 state: completed · document attached · resubmission triggered

A2A in Production · What the Protocol Provides

What A2A delivers that
custom agent wiring cannot.

Capability	How A2A Implements It	Production Significance
Agent discovery	Every PLRX specialist agent publishes an Agent Card at `/.well-known/agent.json` — name, endpoint URL, declared skills, and authentication requirements. Any A2A-compatible client discovers capabilities before contact. No out-of-band configuration. No SDK required.	New agents join the fleet without integration work. The orchestrator discovers them via Agent Card. The protocol is the integration.
Task lifecycle management	Every task submitted via `message/send` moves through a defined lifecycle: `submitted → working → input_required → completed`. Each transition is persisted by the PLRX Durable State Machine. Tasks resubscribe for streaming updates via `tasks/resubscribe`.	Long-running tasks — waiting days for a provider document, a payer response, or a supplier confirmation — suspend durably and resume with full context. No polling loops. No timeout logic. No state loss.
Idempotency enforcement	Every task submission includes a deterministic idempotency key. If the client retries a failed submission, the platform returns the existing task state without executing the operation twice. No duplicate emails, no duplicate API calls, no duplicate submissions.	Exactly-once semantics on every side-effecting operation across the entire agent fleet. Retry-safe by architecture, not by convention.
Specialist agent delegation	The orchestrator mission delegates to specialist agents via A2A: Clinical Documentation Specialist, Billing Service Agent, Data Scientist Agent — each running as an independent A2A server, each returning results to the orchestrator via the same protocol.	Complex multi-party operational workflows decompose into specialist agent tasks. Each specialist is independently upgradeable. The protocol is the coordination contract.
Open protocol — no lock-in	A2A is an open JSON-RPC 2.0 standard. Any A2A-compliant agent — built by PLRX, built by your team, or provided by a third party — participates in the same coordination protocol. Portable across implementations.	Your systems, your integrations, and third-party agents can coordinate with PLRX agents via the same open standard. No proprietary SDK required on either side.

Enterprise Fleet Governance · The Question That Determines Deployability

Can you govern it, suspend it, and audit it independently?

Enterprise AI deployments that run agent fleets across regulated data have one non-negotiable governance requirement: the ability to stop any agent, at any level, immediately — and produce a complete record of everything that happened without vendor involvement.

PLRX answer: three-level suspension, unified audit trail, no vendor call required. Every A2A task delegation — from orchestrator to specialist, specialist result back to orchestrator — is logged in the PLRX WORM audit trail with task ID, method, timestamp, input, and result. The complete mission record, including every inter-agent coordination event, is queryable from a single record by mission ID.

Suspension is immediate at three levels: platform-wide (halt all agents), agent-level (halt a specific specialist agent type), or workflow-level (halt a specific open mission). State is preserved at suspension — the task can be resumed or forensically examined. If a regulatory examination asks for the complete action sequence on a specific mission, the A2A task log provides it — without calling PLRX.

CTO · VP Engineering · A2A Architecture

A2A is not a roadmap item. It is the coordination protocol behind every PLRX mission running in production today.

If your engineering team is evaluating agentic architecture — whether to build on A2A, whether PLRX is the right platform for your operational workflows, or whether to deploy versus build — the scoping call is where that conversation starts. Bring your CTO or VP Engineering.