a2a-splunk

(日本語版 README: README.ja.md)

An independent bridge agent that exposes the Splunk Observability Cloud MCP server over the A2A (Agent2Agent) protocol.

A2A is a protocol for "agents talking to agents as peers"; MCP is a protocol for "agents calling tools" — the two have different design philosophies. This project explores what it takes to bridge them, using the Splunk MCP server (GA'd February 2026) as the concrete example.

---

Architecture

flowchart LR
    SM["Splunk Observability Cloud"]
    SP["Splunk MCP server"]
    BR["splunk_a2a_bridge.py<br/>(A2A server, :8200)"]
    UI["splunk_bridge_ui.py<br/>(investigation chat UI, :8210)"]
    CL["Any A2A client"]

    SM --> SP
    SP <-->|MCP| BR
    CL -->|A2A message:send| BR
    UI -->|A2A message:send| BR

Loading

Files

File	Role
splunk_a2a_bridge.py	A2A server itself, implemented with the a2a-sdk's AgentExecutor pattern
splunk_mcp_tools.py	Splunk MCP server tool-calling logic (binds tools to an LLM and lets it answer on its own)
splunk_bridge_ui.py	A simple investigation chat UI (NiceGUI) for asking the bridge natural-language questions
llm_factory.py / response_schema.py / i18n.py	Shared utilities reused across several of my personal projects (unmodified)
config.ini.example	Configuration file template
telemetry/	Example telemetry-collection config used for verification in a personal lab (containerlab, two Linux hosts + Arista cEOS). Ships metrics to Splunk Observability Cloud via the Splunk OTel Collector (host metrics) and telegraf (gNMI)

Setup

1. Dependencies

pip install a2a-sdk sse_starlette fastapi uvicorn nicegui httpx \
            langchain-mcp-adapters langchain-core langchain-openai

2. Create config.ini

cp config.ini.example config.ini

Set your LLM API key in the [GROQ] section, and your Splunk MCP Gateway connection info in the [SPLUNK] section.

[GROQ]
GROQ_API_KEY = your_groq_api_key_here
GROQ_BASE_URL = https://api.groq.com/openai/v1
GROQ_MODEL    = openai/gpt-oss-120b

[SPLUNK]
SPLUNK_MCP_URL   = https://region-xxxx.api.scs.splunk.com/system/mcp-gateway/v1/
SPLUNK_SF_REALM  = your_realm
SPLUNK_SF_TOKEN  = your_token_here

SPLUNK_MCP_URL varies by realm — check the official docs' Supported Regions page.

3. Run

# Terminal 1: the bridge itself
python splunk_a2a_bridge.py       # :8200

# Terminal 2: investigation UI
export BRIDGE_URL=http://localhost:8200
python splunk_bridge_ui.py        # :8210

Open http://localhost:8210 in your browser and ask a question, either via a suggestion button or free text.

4. Calling it directly, without the UI

curl -X POST http://localhost:8200/message:send \
  -H "Content-Type: application/json" \
  -H "A2A-Version: 1.0" \
  -d '{"message":{"role":"ROLE_USER","messageId":"1","parts":[{"text":"Are there any recent alerts?"}]}}'

License

MIT License (see LICENSE) Copyright (c) 2026 hidemi-k