You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
**Agent registration, discovery, and reliable async messaging — in one broker.**
28
+
---
4
29
5
30
mq9 is the infrastructure layer for multi-agent systems. It solves the two foundational problems every multi-agent system faces: how Agents find each other, and how they communicate reliably and asynchronously.
6
31
7
-
Unlike general-purpose registries (etcd, Consul) combined with general-purpose queues (Kafka, NATS), mq9 is designed specifically for Agent communication. It provides an AgentCard data model, capability-based semantic discovery, per-Agent persistent mailboxes, N-to-N Agent topology, and long-task state retention — all in a single broker with shared runtime, storage, network, and cluster coordination. Agent state and message routing are consistent end-to-end.
32
+
Unlike general-purpose registries (etcd, Consul) combined with general-purpose queues (Kafka, NATS), mq9 is designed specifically for Agent communication. It provides an AgentCard data model, capability-based semantic discovery, per-Agent persistent mailboxes, N-to-N Agent topology, and long-task state retention — all in a single broker with shared runtime, storage, network, and cluster coordination.
33
+
34
+
mq9 natively supports the **A2A (Agent-to-Agent)** protocol via the `mq9.a2a` SDK facade, wrapping the official a2a-sdk so developers can build a fully compliant A2A Agent in 15 lines of code.
35
+
36
+
> **mq9 = Broker + SDKs**
37
+
> The **Broker** (server) code lives in [github.com/robustmq/robustmq](https://github.com/robustmq/robustmq) — mq9 is RobustMQ's fifth native protocol, documented at [robustmq.com](https://robustmq.com).
38
+
> This repo contains the **SDKs** (Python, Java, Go, Rust, JavaScript), LangChain toolkit, demos, and the [mq9.robustmq.com](https://mq9.robustmq.com) documentation site.
8
39
9
-
mq9 natively supports the A2A (Agent-to-Agent) protocol via the `mq9.a2a` SDK facade, wrapping the official a2a-sdk so developers can build a fully compliant A2A Agent in 15 lines of code. Native protocol access and forward compatibility with MCP and ANP are also supported.
AI Agents communicate differently from traditional services. Agent tasks are long-running (LLM inference, multi-step tool calls, human approvals). Agents are frequently offline or started on demand. Agents discover each other by capability, not by fixed address. Multi-agent collaboration requires N-to-N topology — something the request-response model doesn't naturally support.
18
49
19
-
The A2A protocol (originally from Google, now evolving under the Linux Foundation) has become the leading open standard for Agent communication. Its 1.0 spec intentionally leaves discovery, reliable transport, and long-task recovery to the ecosystem — explicitly marking discovery as "outside the scope of the A2A protocol itself" and providing transport extension points for community implementation.
50
+
The A2A protocol has become the leading open standard for Agent communication. Its spec intentionally leaves discovery, reliable transport, and long-task recovery to the ecosystem. A survey of the A2A ecosystem shows the SDK, framework integration, and platform Runtime spaces are crowded — but Agent registry, reliable async transport, and protocol-neutral communication infrastructure are clear gaps.
20
51
21
-
A survey of the A2A ecosystem (awesome-a2a) shows the SDK, framework integration, and platform Runtime spaces are crowded — but Agent registry, reliable async transport, and protocol-neutral communication infrastructure are clear gaps. Teams building multi-agent systems today must either assemble general-purpose tools (etcd + Kafka + glue code) or accept the limitations of HTTP-only communication. Neither is ideal.
22
-
23
-
mq9 fills that gap: a broker purpose-built for Agent communication, with registration and messaging unified in one system.
52
+
Teams building multi-agent systems today must either assemble general-purpose tools (etcd + Kafka + glue code) or accept the limitations of HTTP-only communication. mq9 fills that gap: a broker purpose-built for Agent communication, with registration and messaging unified in one system.
24
53
25
54
---
26
55
27
-
## Two Problems. One Broker.
56
+
## 🤖 Two Problems. One Broker.
28
57
29
58
**Problem 1 — Agents cannot find each other.**
30
59
Agents are dynamic. They come online with different capabilities, at different times. Without a registry, every team hardcodes addresses or builds their own directory. mq9 gives every Agent a place to publish its capabilities and be discovered by others — by keyword or by semantic intent.
31
60
32
61
**Problem 2 — Agents are not always online at the same time.**
33
62
Agents are task-driven — they start, execute, and stop. When Agent A sends to Agent B and B is offline, the message is lost. HTTP requires both sides to be online. Redis pub/sub has no persistence. mq9 solves this with persistent mailboxes: send a message, the recipient gets it whenever they come online.
34
63
64
+
| Problem | mq9's answer |
65
+
|---------|-------------|
66
+
| How do Agents find each other? | Built-in registry: `AGENT.REGISTER` + `AGENT.DISCOVER` with full-text and semantic vector search |
67
+
| How do Agents communicate reliably? | Persistent mailbox per Agent: messages wait until the recipient is online, with 3-tier priority and FETCH+ACK pull consumption |
10 commands over NATS request/reply under `$mq9.AI.*`:
170
+
mq9 natively supports the **A2A (Agent-to-Agent)** protocol. The `mq9.a2a` SDK facade wraps the official `a2a-sdk` — build a fully compliant A2A Agent in 15 lines of code, with mq9 handling discovery, persistent delivery, and priority routing underneath.
152
171
153
-
| Category | Subject | Description |
154
-
| -------- | ------- | ----------- |
155
-
| Registry |`$mq9.AI.AGENT.REGISTER`| Register Agent with capability description |
0 commit comments