Bewize Request Demo

Hermes Hub release notes

The control plane is turning agents into operated infrastructure

A chronological view of Hermes Hub progress: tenant-scoped runtimes, A2A edge access, redacted observability, policy, release management, secrets, storage, schedules, flavors, and production operator surfaces.

Updated July 9, 2026

Explore Hermes HubWorkCord releases
Tenants isolated runtime boundary
A2A controlled agent edge
Cron scheduled agent work
Metrics redacted usage visibility

Chronological changelog

What changed, latest first

These notes focus on source-backed product capabilities and avoid overstating roadmap items as shipped.

Current platform inventory

Latest

Hermes Hub product truth was consolidated into an operator-ready control-plane story.

The current Hermes Hub inventory frames the product as the operational control plane behind hosted agents: isolated by tenant, governed by policy, observable without leaking prompts, connected to chat and A2A, upgraded through releases, packaged through flavors, and kept cost-efficient through cold-idle hosting.
Isolation Tenant boundaries are concrete infrastructure

Each tenant has isolated memory, secrets, sessions, cron, files, browser profile, SSH identity, logs, and runtime state. The public claim is infrastructure separation, not a prompt promise.

Operations Runs, releases, and schedules are managed surfaces

The Hub exposes tenant lifecycle, run status/events, stop controls, cron, release activation, pinned releases, reconcile flows, and timeline markers so agent work can be operated like a product.

Governance Capabilities are granted, not assumed

Hub-wide and per-tenant policy controls search, files, browser, terminal, cron, delegation, and similar features; approval gates exist for sensitive actions.

  • Boundary: current shipped runtime scope is Hermes Agent tenants. Arbitrary agent-framework hosting and a public third-party flavor marketplace should not be claimed as shipped.

Source: Hermes Hub maintained feature inventory and article research brief refreshed 2026-07-09.

Admin proof

Hermes Hub admin UI gained screenshot-backed tenant operations proof.

The Hermes Hub Playwright output demonstrated dashboard health, tenant lifecycle controls, tenant-scoped timeline observability, policy and skills management, secrets/env redaction, system operations, browser isolation, and hub config defaults with tenant runtime impact.
Console Operators can work without SSH as the default interface

Tenant directory/detail, lifecycle controls, policy and skills views, file browsing, schedule markers, env metadata, releases, specialization, reconcile output, and browser-isolation proof became screenshot-ready surfaces.

Trust Secrets and metrics stay product-safe

Secret metadata and usage surfaces are designed to support operations without rendering raw secret values or turning analytics into a prompt leak.

Source: 2026-06-14 product-update feed.

Website architecture

Bewize public website started presenting the A2A ecosystem product architecture.

The website began leading with the Bewize A2A ecosystem: WorkCord for shared rooms, Hermes Hub for hosted agent runtime and orchestration, AgentsEval for captured-run evaluation, and Browser Sidecar for tenant browser isolation. The copy separated implemented proof from directional Chrome-extension and process-mining claims.
Ecosystem Hermes Hub became the platform layer in the public story

The website connects WorkCord, Hermes Hub, AgentsEval, and Browser Sidecar as one managed A2A agent ecosystem rather than isolated product pages.

Positioning Private and governed deployment became the center

The story moved toward self-hosted, isolated, observable agent infrastructure for companies that need control over data, tools, runtime policy, and adoption.

Source: 2026-06-14 product-update feed.