Urgent.News

What's breaking now, across thousands of outlets.

Editions

AI

Grok, Claude, and Hermes agents get job titles — and persistent permissions

Your next AI coworker may look like a chatbot, but underneath their name, face, and job title will be something The post Grok, Claude, and Hermes agents get job titles — and persistent permissions appeared first on The New Stack .

Grok, Claude, and Hermes agents get job titles — and persistent permissions

AI-powered agents such as Grok, Claude, and Hermes are now being assigned job titles and persistent permissions, marking a significant shift in their capabilities. This development is evident in the Hermes desktop app, which has incorporated Bot Mode into its core functionality. The standalone plugin repository, previously archived, is now bundled and enabled by default, with v0.20.4 introducing a tabbed Sessions and Bots sidebar.

Each profile for these agents appears as a named bot, complete with its own chat, avatar, personality, schedule, and pinned model. This reflects a broader trend in which sessions have been integrated alongside bots, maintaining their ability to carry memory and permissions across conversations. Hermes and Grok build rosters, while Claude joins a roster shared by xAI, offering a different approach to organization.

Claude Tag from Anthropic, introduced on June 23, takes a unique route by allowing one organization-provisioned Claude to join the human roster in Slack. Each thread still initiates its own working session. Across these three products and deployment models, a reusable object has transitioned from being the conversation to serving as the worker identity above it. A bot, in this context, is a profile rather than a new runtime environment.

Hermes provides a clear example of this transformation. A bot is a Hermes Agent profile, complete with a separate home directory holding configuration, keys, a SOUL file, memory, sessions, skills, and cron state. Routines, which are ordinary cron jobs, are namespaced per bot. Bot-to-bot handoffs are facilitated through real command-line interface (CLI) invocations into a canonical bot chat, triggered whenever the receiving bot next runs.

The standalone plugin acts as a user interface over these primitives, with no core patches or background daemons. Profiles have existed prior to Bot Mode, which merely added faces and a sidebar to these profiles.

The shift in identity architecture is significant. Descriptions now carry real weight in these products, enabling vendors to encode a bot’s durable responsibilities, sources, and boundaries in their description. Bots can then hand off work to one another, with descriptions acting somewhat like a routing table. Across these three products, three distinct identity architectures have emerged: Hermes keeps credentials profile-scoped to the host running Hermes; Grok's approach shares one account-scoped computer for all bots; and Claude Tag provisions an organization service identity with channel-scoped access to tools.

Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at thenewstack.io →

More in AI

More from Friday 21 August →