Problem
agent-relay node agent attach and the message subcommands accept --broker-url, --api-key, and --state-dir, but agent-relay node agent release <name> accepts none of them.
During a fleet drain from a directory without .agentworkforce/relay/connection.json, release failed with:
No running broker found (.../.agentworkforce/relay/connection.json does not exist)
Setting RELAY_BROKER_URL and RELAY_BROKER_API_KEY did not change the lookup path in the released CLI.
Expected
Release should support the same explicit local-broker selection contract as attach and message:
agent-relay node agent release worker --state-dir ~/.agentworkforce/relay/sf-mini-node/state
agent-relay node agent release worker --broker-url URL --api-key KEY
It should also have coverage proving explicit selection works outside the broker project directory.
Why it matters
Fleet nodes intentionally keep isolated state directories. Requiring the operator to run from a synthetic project connection path makes safe drains brittle and forced this incident response to use the broker HTTP API directly.
Problem
agent-relay node agent attachand the message subcommands accept--broker-url,--api-key, and--state-dir, butagent-relay node agent release <name>accepts none of them.During a fleet drain from a directory without
.agentworkforce/relay/connection.json, release failed with:Setting
RELAY_BROKER_URLandRELAY_BROKER_API_KEYdid not change the lookup path in the released CLI.Expected
Release should support the same explicit local-broker selection contract as attach and message:
It should also have coverage proving explicit selection works outside the broker project directory.
Why it matters
Fleet nodes intentionally keep isolated state directories. Requiring the operator to run from a synthetic project connection path makes safe drains brittle and forced this incident response to use the broker HTTP API directly.