Knowledgebase

dpanel.ai Knowledgebase

Practical guides for AI Execution Deployment Panels, deployable AI teams, approvals, grounding, handoff, integrations, deployment, and governance.

21 articles

Abstract dpanel.ai control panel for AI team execution
Fundamentals · 5 min read

What is an AI Execution Deployment Panel?

An AI Execution Deployment Panel is the operator layer for deploying AI teams that can be assigned work, grounded in context, approved by humans, and connected to real tools.

AI workers grouped into a managed team workflow
Fundamentals · 6 min read

AI Teams vs AI Agents

AI agents can perform tasks. AI teams need roles, ownership, approvals, handoff, and visibility so a business can actually operate them.

Human review step before an AI workflow writes to external tools
Operating AI Teams · 7 min read

Human Approval Gates for AI Workflows

Approval gates keep risky AI actions behind human review before external writes, customer-visible messages, spend, deployment, or status changes happen.

Ledger view of AI team context, tasks, and approved changes
Operating AI Teams · 5 min read

Grounding Ledger Explained

A grounding ledger records what context an AI team used, what changed, and why a worker acted the way it did.

Structured handoff packet between AI worker, human, and external tool
Operating AI Teams · 6 min read

Handoff Packets for AI Teams

A handoff packet turns AI output into an operational transfer: owner, status, risk, blockers, links, evidence, and next action.

monday.com board connected to an approval-aware AI team workflow
Integration Guides · 8 min read

How dpanel.ai Works Around monday.com

monday.com can stay the board of record while dpanel.ai manages AI-team execution, approvals, handoff, and approved status updates around it.

DPanel Pack and BYOK deployment paths for AI teams
Deployment · 5 min read

DPanel Pack vs BYOK

DPanel Pack is the optional managed model-usage path with prepaid credits and metered calls. BYOK stays the default path and DPanel never charges BYOK customers for model usage.

Prepaid credit ledger for DPanel Pack model usage
Deployment · 5 min read

How Pack Billing Works: Credits, Holds, and Margin

Pack billing uses prepaid credits, pre-flight holds, metered model calls, and transparent margin so customers can control managed AI-team spend.

BYOK provider key applied to a customer-owned DPanel instance
Deployment · 4 min read

Use Your Provider Key on a DPanel Instance

Apply a connected BYOK provider key to one of your own DPanel instances with explicit consent. The key goes into your instance only and DPanel never charges BYOK model usage.

Deployment target choices for DPanel AI-team instances
Deployment · 5 min read

Where Can My DPanel Instances Run?

DPanel is the control plane, not the hosting company. Instances can run locally, on an existing remote host, or on supported cloud and partner targets as they become available.

Server-side guardrails applied before AI model calls
Security & Governance · 6 min read

Guardrails Overview

DPanel Guardrails apply usage and safety policy server-side before every model call, with a workspace default and per-instance override model.

Budget policy caps for AI-team usage
Security & Governance · 3 min read

Budget Policies

Budget Policies set monthly and daily spend caps plus request- and token-per-minute rate caps for DPanel model calls.

Allowed and denied model provider policy for an AI-team instance
Security & Governance · 3 min read

Model & Provider Access

Model & Provider Access policies allow or deny the model and provider paths each DPanel instance may use.

Prompt injection attempt blocked before a model call
Security & Governance · 3 min read

Prompt Injection Guardrails

Prompt Injection guardrails detect jailbreak and instruction-override attempts before model calls run.

Sensitive data detection policy for AI-team prompts
Security & Governance · 3 min read

Sensitive Info Detection

Sensitive Info Detection identifies PII, credentials, and personal data in prompts so DPanel can flag, redact, or block the request.

Isolated dpanel Box runtime versus an agent with full workstation access
Security & Governance · 7 min read

Why Dedicated Boxes Beat Agents on Your Laptop

Running an agent framework on a workstation hands it your SSH keys, browser sessions, and VPN into the whole organization. A dpanel Box contains the blast radius to one isolated runtime with only the credentials you chose to give it.

Separate isolated Boxes per client entity with their own tools and vaults
Security & Governance · 5 min read

One Box per Client: Entity Separation with Multi-Box

Multiple Boxes let an operator give every client, brand, or business entity its own isolated runtime — separate git identities, separate monday.com boards, separate Notion workspaces, separate vaults — with zero cross-contamination.

dpanel.ai runtime transparency and Hermes Agent attribution
Deployment · 6 min read

Why dpanel.ai Discloses Hermes Agent

dpanel.ai runs on a hardened fork of Hermes Agent by Nous Research under MIT terms, then adds the control plane for teams, approvals, grounding, usage, and handoff.

DPanel board metrics and status vocabulary explained
Glossary & Concepts · 9 min read

Panel Terms Explained: Drift, QA Gate, Judge, and Friends

Plain-English definitions for every status, metric, and chip the DPanel operator surfaces show — what each one measures, where it appears, and when it changes.

The per-agent document set: persona, instructions, and persistent memory
Glossary & Concepts · 8 min read

SOUL.md, AGENTS.md, MEMORY.md: The Agent’s Files, Explained

The document tabs on a team member are real files in the agent’s runtime home — the same conventions the Hermes Agent runtime documents. What each file is, who writes it, and how DPanel keeps them synced.

The DPanel left menu vocabulary explained
Glossary & Concepts · 6 min read

Workspace, Loops, Co-workers, Earnings Center: The Menu, Explained

The NAV V2 vocabulary — what a Workspace is (and the master workspace), why scheduled tasks are called Loops, who Co-workers are, and what the Earnings center pays for.