Skip to main content
An agent runtime speaks ACP to its client, calls tools over MCP, and fetches grounding context over RCP. All three share JSON-RPC framing, an initialize/capabilities handshake, _meta extensibility, and cursor pagination — learn one, you know the shape of all three.

​
Roles

Client

An application, agent runtime, or IDE that needs retrieval context. It initiates the connection and drives the handshake.

Server (engine)

A RAG engine, vector database, or aggregator that advertises capabilities and answers method calls.
The client always initiates. A server MUST NOT assume any capability was negotiated until initialize completes.

​
The retrieval pipeline model

RCP models retrieval as a composable pipeline. A single retrieve call can transparently drive as many stages as the engine supports:
1

Query transform

Expansion, HyDE, decomposition (query/transform).
2

Candidate generation

Dense, sparse, or hybrid first-stage recall.
3

Rerank

Cross-encoder or late-interaction (MaxSim) scoring.
4

Diversify

MMR to reduce redundancy.
5

Assemble

Apply metadata filters and recency, and attach citations.
A thin client just calls retrieve and reads hits; a sophisticated client can also call the individual stages (embed, rerank, query/transform, graph) directly. See The retrieval pipeline for the full model.

​
Message shape

Every message is JSON-RPC 2.0. Requests carry an id; notifications do not. Extensibility rides in _meta.

Message format

The normative framing, id rules, and _meta extensibility.

​
How the connection unfolds

1

Connect

The client spawns the server (stdio) or opens an HTTP endpoint.
2

Initialize

The client sends initialize with its protocol version; the server replies with the negotiated version, its identity, and its capabilities.
3

Call

The client calls capability-gated methods (retrieve, embed, …), reading results and optionally streaming progress.
4

Shut down

The client sends shutdown (or just closes the transport).