Start here
Sentinel documentation
Build, connect, orchestrate, and operate AI agents from one observable workspace.
What Sentinel is
Sentinel is a business-facing agent management platform. Teams configure model-powered agents, ground them with private knowledge, connect tools through MCP, compose multi-agent fleets, and inspect every run from input to output.
Agents
Independently testable workers with a model, instructions, knowledge, tools, memory, and optional structured output.
Fleets
Visual, durable workflows that coordinate agents, decisions, parallel work, approvals, and external calls.
Grounded knowledge
Indexed documents that agents retrieve and cite instead of guessing from model memory.
Observable runs
Status, timing, token usage, retrieval hits, tool calls, and trace detail for every execution.
How work moves
Input
Question and context
Agent or fleet
Reason, retrieve, act
Run trace
Output and evidence
Choose your path
New to Sentinel
Follow Quickstart, then use the Create an agent example with one representative task.
Designing a workflow
Learn Agents, Tools & MCP, and Fleets before building the Fleet walkthrough.
Integrating an application
Read API integration after testing and deploying the agent or fleet you want to invoke.
Operating production workloads
Focus on Runs, security boundaries, budgets, approvals, and failure behavior.