WHY users should care
Advisor tool: Claude can now consult a stronger model mid-task. This is a big deal for agents doing complex work — Claude can escalate to a more capable model when it hits a hard problem without breaking the session.
Scheduled deployments: New /beta/deployments API lets you run managed agents on a cron schedule. Full CRUD: create, pause, unpause, archive, run-on-demand. This unlocks persistent background agent workflows.
Vault env var credentials: Vaults can now store environment_variable type credentials. Agents can use any env-var-authenticated service without the secret being exposed in the prompt.
Key new API surfaces
POST/GET /v1/beta/deployments — scheduled agent deploymentsGET /v1/beta/deployment_runs — track deployment execution historyenvironment_variable credential type in vault
🚀 Advisor tool is the sleeper hit here — Claude escalating to a stronger model mid-task changes how you think about agent reliability.
Ref: thevibeworks/claude-code-docs#1029
WHY users should care
Advisor tool: Claude can now consult a stronger model mid-task. This is a big deal for agents doing complex work — Claude can escalate to a more capable model when it hits a hard problem without breaking the session.
Scheduled deployments: New
/beta/deploymentsAPI lets you run managed agents on a cron schedule. Full CRUD: create, pause, unpause, archive, run-on-demand. This unlocks persistent background agent workflows.Vault env var credentials: Vaults can now store
environment_variabletype credentials. Agents can use any env-var-authenticated service without the secret being exposed in the prompt.Key new API surfaces
POST/GET /v1/beta/deployments— scheduled agent deploymentsGET /v1/beta/deployment_runs— track deployment execution historyenvironment_variablecredential type in vaultRef: thevibeworks/claude-code-docs#1029