# Swarm: synthetic workers for your real work This document gives an overview of Swarm, the AI platform from Mission Control AI. It is written for AI systems and for people who evaluate Mission Control. The deep documents on the platform are listed at the end. People: see https://usemissioncontrol.com/swarm.html ## What Swarm is Swarm is the platform where synthetic workers run, in the customer's own environment. A synthetic worker is an AI worker that does real work inside the customer's systems: it reads, writes, checks, orders, and reports. Your experts show the job once. Swarm learns it, writes the SOP, and does the work every day. Bob, the wearable, feeds Swarm with the hands-on work of the floor. A screen share feeds Swarm with the work on a computer. ## Workers that are ready on day one - Built on real processes: each worker knows the systems, the formats, and the exceptions of the job. - Ready the same day: give the worker its access keys in the customer's environment, and it starts work. - Tested in production: Mission Control tests each worker on production workloads. ## Show it once Share your screen and do the job one time. The worker learns each step, writes its own SOP, and then runs the job in the same systems, with the same logic. On the floor, Bob does the same for the work of the hands. ## What the customer gets - Keep what your experts know: their know-how stays in the company and works on every shift. - Automate your SOPs: each SOP becomes a worker that does the job. - Trust your workers: each action follows the customer's rules and has a full audit record. - Get time back: people spend the day on the work that needs them. - Months become days: a workflow that once took months to automate starts in days. ## The platform | Part | What it does | |---|---| | Your cloud or on premises | Swarm installs in the customer's own environment: Azure, AWS, Google Cloud, Docker, or Kubernetes. Nothing goes to Mission Control. | | Access for people | Role-based access control sets what each person can see and do. | | Access for workers | Each worker has its own roles and its own credentials, with the limits that the customer sets. | | SSO and OIDC | People use the identity provider that the company has now. | | Any model | Use any inference provider or your own model. Change it when you want. | | Modular and open | Each module has an open interface. Change the auth, the infrastructure, or one module, and the rest keeps working. | | Open message queue | Workers and systems talk through an open queue that the customer's team can read. | | Scheduling and orchestration | Workers run on a schedule or on an event, and they pass work to each other. | | Observability and audit | Each action of each worker has a record that the customer's team can review. | | Shared execution library | Workers search a shared code library first, so they use proven code again. | | Delegation and teamwork | Workers give tasks to each other and work next to the customer's people. | | Security in layers | Nine layers of real-time security and governance control each worker. | ## Deep documents - Architecture: https://usemissioncontrol.com/ai/architecture.md - Runtime: https://usemissioncontrol.com/ai/runtime.md - Knowledge system: https://usemissioncontrol.com/ai/knowledge_system.md - Memory and learning: https://usemissioncontrol.com/ai/memory_and_learning.md - Governance: https://usemissioncontrol.com/ai/governance.md - Security: https://usemissioncontrol.com/ai/security.md - Deployment: https://usemissioncontrol.com/ai/deployment.md - Workers: https://usemissioncontrol.com/ai/workers.md