01 / DECIDE
Architecture decision
Use it to add live external tools or context without hard-coding each integration into the agent.
Record server identity, transport, authentication, tool namespace, enabled tools, approvals, and diagnostic behavior. Map input, output, state, tools, and side effects as one observable path, then name the owner of connect mcp servers with a narrow trust contract.
02 / OPERATE
Three-step field procedure
- 01
Lock the contract
Write down the caller, data classification, success condition, timeout, cancellation, and ownership. Use it to add live external tools or context without hard-coding each integration into the agent.
- 02
Run one bounded path
Build one end-to-end path with correlation, typed state, and reversible failure handling. Record server identity, transport, authentication, tool namespace, enabled tools, approvals, and diagnostic behavior.
- 03
Verify the signal
Turn acceptance into a repeatable fixture, contract test, or browser test. Doctor and negative tests verify identity, expired auth, schema drift, unreachable transport, and injected content.
03 / BOUND
Operational boundary
Connect only trusted servers, keep credentials outside committed config, and minimize the exposed tool set.
04 / PROVE
Acceptance evidence
Doctor and negative tests verify identity, expired auth, schema drift, unreachable transport, and injected content.
SOURCE / HTTP
Reproducible source probe
curl -fsSI 'https://docs.x.ai/build/features/mcp-servers' | sed -n '1,5p'