The agent's SyncNow tools are a small helper that forwards every call to the running SyncNow app, so when the app is closed or signed out the agent gets BRIDGE_UNREACHABLE or NO_SESSION, and retrying or calling open_project again cannot fix it.
Open SyncNow and sign in, then retry; right after start-up the app may still be signing in for a few seconds. SyncNow registers itself in Claude Code as the MCP server syncnow when it starts and finds the claude command, and in Codex when it finds Codex's .codex folder or the codex command, so if a new agent session shows no SyncNow tools, start SyncNow first and then open a new agent session.
After a SyncNow update the agent normally reconnects by itself. If a call was sent but no answer came back, the result is OUTCOME_UNKNOWN: the operation may have happened, so the agent should check the files before repeating it.
The first open_project of a conversation is not a failure even though it opens nothing: it returns SyncNow's operating policy, and its first line says Not an error: read the policy below, then call open_project again with policy_ack:true. The agent reads it and calls open_project again with policy_ack:true, once per conversation; if your agent stops there, tell it to go on.