# MACH Agent Lab Edge

Independent LAB_ONLY discovery and interoperability research surface for autonomous agents.

IMPORTANT
- This is not MACH Base production.
- Production connected: false.
- Payments/wallets/credentials: none.
- Production execution: disabled.
- All executable capabilities are deterministic lab.mock.* fixtures with no side effects.

DISCOVERY
MCP endpoint: POST /api/mcp
MCP Server Card: GET /api/mcp/server-card
Agent Card: GET /.well-known/agent-card.json
ARD: GET /.well-known/ard.json
Capabilities: GET /api/capabilities
Action contract: GET /actions.json
Direct discovery: GET /api/discover?q=<natural-language-intent>

A2A 1.0
JSONRPC endpoint: POST /api/jsonrpc
HTTP+JSON endpoint: POST /api/a2a/message:send
Task list: GET /api/a2a/tasks
Input role: ROLE_USER
Output role: ROLE_AGENT
Streaming: false
Push notifications: false
Tasks created by SendMessage: false; direct Message responses only.

DETERMINISTIC INVOCATION
Send a data Part with:
{"skillId":"lab.mock.wallet-format","input":{"address":"0x0000000000000000000000000000000000000000"}}

SAFETY
POST /api/execute always returns HTTP 503 production_execution_disabled.
Do not send secrets, credentials, private customer data, wallet keys or payment proofs.

Use this edge to test capability discovery, selection, interoperability and harmless deterministic execution. Do not infer that a lab.mock.* fixture is a live MACH Base service.