Salesforce iconSalesforceSep 28, 2026 ~7 min source read

NVIDIA OpenShell and Slack bring runtime admin approvals into the flow of work

OpenShell enforces runtime policies for AI agents; the Slack Admin Bridge surfaces access exceptions and approvals inside Slack so admins can decide, discuss, and audit agent requests without leaving their workflow.

Extending Admin Control to the Runtime Layer with NVIDIA OpenShell and Slack

Share this story

Send the public story page.

Useful takeaways from this story.

The Slack Admin Bridge delivers approval cards and real-time analytics into Slack while keeping credentials and message data out of the bridge.

The integration runs inside the customer’s environment with permissions scoped to send audit events and ingest approvals only.

Salesforce and NVIDIA plan further exploration of combining OpenShell with MuleSoft Agent Fabric for broader governance and runtime enforcement.

# What this announcement means

# How the flow works in practice

An agent running inside OpenShell requests access to a resource outside its policy boundary. OpenShell blocks or flags the request according to runtime policy. When an exception occurs, the Slack Admin Bridge creates an approval card that appears in Slack for designated admins and channels. Admins can review the request, approve or deny it, and carry on the approval discussion within the same Slack thread. OpenShell receives the approval decision and continues execution according to policy.

# Security and separation of concerns

The integration separates three responsibilities: the agent that asks for access, OpenShell that enforces policy at runtime, and humans in Slack who authorize exceptions. Key security points described in the announcement:

  • OpenShell credentials and sensitive runtime secrets remain outside Slack.
  • The bridge runs inside the customer's secure environment and connects to Slack with narrowly scoped permissions.

That design limits the bridge's surface area to approval workflow and audit telemetry rather than becoming a conduit for broader data movement.

# Admin experience and observability

Admins receive interactive approval cards directly in Slack, so they don't need to leave their normal workflow to make decisions. The bridge also surfaces a real-time analytics view in Slack's App Home, showing outstanding requests, past decisions, and agent activity. The conversation, debate, and audit trail all live in Slack, tying context and approvals together where teams already collaborate.

# Availability and setup

According to the announcement, the OpenShell + Slack integration is available for developers today. Once OpenShell is deployed, developers can connect the Slack Admin Bridge and configure which admins and channels receive approval requests. The bridge runs in customers' environments and uses scoped permissions to limit what it can send and receive.

# Where this could be used next

Salesforce and NVIDIA say they are exploring an integration path that pairs OpenShell's runtime enforcement with MuleSoft Agent Fabric's governance and orchestration. That combination would aim to govern which agents can access systems and enforce what those agents are allowed to do while they run.

# Practical takeaway

If your organization is deploying longer-running, multi-system AI agents, OpenShell plus the Slack Admin Bridge offers a concrete model for runtime enforcement plus human-in-the-loop approvals inside Slack. The architecture keeps credentials and sensitive data out of Slack, centralizes approvals and audit trails in a familiar collaboration tool, and preserves runtime control inside OpenShell.

More context around this story.

NVIDIA Launches Open Agent Safety Platform: OpenShell Sandboxes Agents on Vera CPUs While Sentry on BlueField-4 Quarantines Them in Milliseconds
Marktechpost iconMarktechpostSep 28, 2026

NVIDIA Launches Open Agent Safety Platform: OpenShell Sandboxes Agents on Vera CPUs While Sentry on BlueField-4 Quarantines Them in Milliseconds

NVIDIA has launched the Open Agent Safety Platform, an open reference design that enforces AI agent safety outside the agent itself. OpenShell, an Apache 2.0 runtime, sandboxes agents under YAML policies. Sentry, an out-of-band watchdog on BlueField-4 DPUs, can quarantine an agent that escapes its boundary in milliseco

Extending public sector intelligence with Agentforce and AWS
Amazon iconAmazonSep 22, 2026

Extending public sector intelligence with Agentforce and AWS

Public sector agencies process large volumes of unstructured evidence, such as body camera footage and scanned documents. This post shows how to combine Amazon Bedrock Data Automation with the Model Context Protocol (MCP) to turn that data into structured insights and surface them through natural language queries in Sa

Loading more related stories...

Keep reading in the app

Open the app view to save this story, compare related coverage, and continue from the same source.

Open in app