AI Agent Hub
Back to plugins
🖥️

dsh-a2a

Client Updated 2026.09.11

Run the following command in DeepSeek Harness:

dsh plugin install HanphoneJan/dsh-a2a

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install HanphoneJan/dsh-a2a in your DeepSeek Harness terminal to install this plugin; source is available at https://github.com/HanphoneJan/dsh-a2a .

About this plugin

When a DeepSeek Harness profile must talk to multiple external agent systems concurrently, traditional A2A integration means hand-editing config files, hard-coding endpoints per peer, and managing authentication one by one — every new connection is another code change. dsh-a2a collapses all of that into a pure GUI workflow: pick a preset, and an inbound server goes live instantly, with its AgentCard and invocable skills derived automatically. No typed forms, no YAML edits.

The plugin implements the full A2A v1.0.1 protocol surface — SendMessage, SendStreamingMessage, complete task lifecycle, and SSE streaming with catch-up frames — and supports multiple concurrent inbound instances (each with its own endpoint, preset, and auth) alongside multiple outbound connections (remote skills auto-mapped to local model tools). Tasks persist in the a2a storage domain and survive restarts. Inbound traffic passes through a governed waterfall where policy plugins can veto or audit. Every contextId maps to a trackable, cancelable DSH session, and the dashboard shows live peer status and task counts in real time.

It is built for developers who need a Harness profile to act as an agent-to-agent hub: exposing agent capabilities upstream while invoking remote skills downstream, with the full instance lifecycle — create, start, stop, edit, remove — handled entirely in the browser. No patches, no shell scripts. The protocol layer, task store, and executor are independent implementations that add no extra intrusion to the existing Harness dependency graph.

Use Cases

  • Expose multiple preset-bound inbound A2A servers from a single Harness profile simultaneously
  • Auto-map remote A2A agent skills as local model tools and route them through the executor
  • Create, start, stop, edit, and remove inbound and outbound A2A instances entirely from the browser dashboard without touching config files

Best For

  • Developers building multi-agent interoperability systems on DeepSeek Harness
  • Platform teams needing A2A protocol integration without per-instance manual configuration
  • Independent builders who want GUI-driven A2A server management with zero scripting