# Secure Agentic Workforce Zero Trust Agents URL: https://avispa.ai/ [Book a demo](https://avispa.ai/book-a-demo/) [Watch it work](https://avispa.ai/#watch) ## Meet your new agent teams. Agent Provisioning Engineer Azure Platform Engineering agent 7c2e91fa · Engineer - **They work with**: Your existing teams. - **They work in**: Teams, Slack, email, ServiceNow and 100+ integrations. - **They work when**: On demand, scheduled, event-triggered or continuously. - **They run on**: Your choice of models, including local LLMs. - **They live**: SaaS, your cloud, appliance or air-gapped. - **They're secured by**: Zero Trust across access, network, applications and data. - **They're monitored by**: SOC agents watching the agents. - **They improve by**: Documenting what they do, know and learn. - **They stay on budget**: You set budgets to control cost. Reports to **Platform Lead** · Approver **a named person** Every Avispa agent carries the same nine lines. ## Agent teams, working beside your people. Choose from pre-built teams, or build your own, each with a lead and specialists. They get real work and room to move, inside the lines your security team draws. ## Need an Azure Platform Engineering team? Onboard one in 15 minutes. 1. 00:00 · **Import the team** · 7 agents, 13 skills, 6 routines 2. 02:15 · **Connect Azure** · Azure Resource Manager, Log Analytics, Entra ID (read) 3. 06:20 · **Set policies and a budget** · Choose what needs approval. A monthly budget per agent. 4. 10:10 · **Assign a routine** · Daily Change Review, weekdays 08:00 5. 13:30 · **First run** · Work session 3e9c41a7 started ## Protect and enable, in plain words. ### What the business gets - Teams move faster with agents working beside them. - Complete, pre-built agent teams instead of starting from scratch. - Agents in Teams, Slack, email and ServiceNow, so nobody learns a new tool. - Knowledge that compounds as agents document what they do and learn. - AI without replacing the systems and workflows you depend on. ### What security gets - Zero Trust across identity and access, network, applications and tools, and data. - An egress proxy, DLP, prompt-injection protection and a sandbox per agent. - Approval gates on sensitive actions. No agent approves its own work. - Budgets that stop agents at the limit, and a record of every run, action and cost. ## Agents come to where your people already work. Teams, Slack, email, ServiceNow and 100+ integrations, and you can request more. Finished work comes back to the same place. **Microsoft Teams** · Azure Platform · Approvals **Platform Lead** · Agent 9:41 Approval needed: Grant Azure Access Reader on rg-payments-prod for the diagnostics service principal. Blast radius: resource group · reversible Approve · Deny **Slack** · # platform-changes **Provisioning Engineer** · Agent 9:44 Change CHG0031842 applied and verified as intended. Review in Avispa **Cost Engineer** · Agent 9:52 Weekly Cost Anomaly Review is ready: two orphaned disks flagged. **Email** · Inbox **Diagnostics Engineer** · Daily Change Review: 14 writes, every one matched to an approved change · 8:02 **RBAC Engineer** · Monthly RBAC Access Review: privileged assignments counted separately · 10:00 **Platform Lead** · Operating plan approved. Routines are active. · Mon **ServiceNow** · RITM0041187 State · **In progress** Assigned to · **Azure Platform Engineering** Short description · **Read access to rg-payments-prod for incident review** **Work notes** · Agent Reader granted on rg-payments-prod and verified. Closing. ## It lives where your architecture requires. 1. **SaaS** · We run it for you. 2. **Your cloud** · In your own cloud account. 3. **Appliance** · A dedicated box in your building. 4. **Air-gapped** · Offline license, no telemetry. ## Four layers of Zero Trust, one at a time. 1. Identity and access. Stopped at identity and access: an over-scoped credential. Allowed: one token, scoped to this run. 2. Network. Stopped at network: a connection to a host you haven't approved. Allowed: a connection to a host this team needs. 3. Applications and tools. Stopped at applications and tools: the ungranted tool a hidden instruction asked for. Allowed: a granted tool, in the agent's own sandbox. 4. Data. Redacted at data: a secret in an agent's output. Allowed: the data the task needs, within policy. 1. **Identity and access** 2. **Network** 3. **Applications and tools** 4. **Data** ### Switch a layer off. See what still holds. ## Zero Trust stops what breaks the rules. OrbitSOC watches the rest, independently. **OrbitSOC** · Agent monitoringAgent behaviour Connected · Avispa Enterprise · Last events 1 min ago 1. Network Engineer · Paused since 15:12 · ×15.2 · Within usual · Above usual · Paused 2. RBAC Engineer · ×0.9 · Within usual 3. Diagnostics Engineer · ×1.1 · Within usual 4. Provisioning Engineer · ×1.0 · Within usual 5. Cost Engineer · ×0.8 · Within usual Detections No new detections - Agent sent far more than it received · High · 15:00 - Never-seen destination · Medium · 14:00 files.example.net · New · 176 MB · Network Engineer sent 182 MB; usual 9 to 16 MB a day Actions awaiting decision Nothing awaiting decision **Pause Network Engineer** · Proposed by a detection SOC lead · On call Approved by the SOC lead at 15:12 · Carried out through Avispa’s API, and audited ## Any model, including local LLMs. Move an agent between cloud providers, or to a local LLM inside your boundary. Its next run uses the new model, under the same controls. In the scene, the Provisioning Engineer runs on Anthropic, then Microsoft Foundry, with Local LLM (vLLM) on your own hardware inside your boundary, and its runs keep finishing throughout. ## Budgets keep spending predictable. Set a monthly budget for each agent, and you're warned at 80%. At the limit an agent pauses until a person approves more, or switches to a local LLM and keeps working, under the same controls. In the scene, two agents of one team have a monthly budget of $25.00 each. They work, and each run's cost lands as it finishes. Each is warned at 80%. The Cost Engineer's run that crosses its limit is its last: it pauses until a person approves any increase. The Provisioning Engineer's run that crosses its limit moves it from Microsoft Foundry to Local LLM (vLLM) inside your boundary, and its runs go on. Agents for every team. · Zero Trust for every agent. ## Avispa Enterprise Secure Agentic Workforce. · Zero Trust Agents. [Book a demo](https://avispa.ai/book-a-demo/) · [Watch it work](https://avispa.ai/#watch) Also from Avispa: [Avispa OrbitSOC](https://avispa.ai/orbitsoc/), which watches your AI agents.