Concepts
This page maps every part of autobotAI to the job it does, so you know which one to reach for. Each row links to its full guide.
Connect
| Component | What it is | Use it when |
|---|
| Integrations | A connection to one of your tools, with its own credentials and scope | You want agents or workflows to read from or act in that tool |
| Linux and Kubernetes agents | Small agents you run inside private networks | Targets are not reachable from the workspace; the agent connects outbound over HTTPS |
| Generic REST API | A connection to any REST API | There is no native integration yet |
| AI models | Bedrock, Azure AI Foundry, OpenAI, OpenRouter or a self-hosted model | Any AI node, agent or Optimus needs a model |
Start work
| Component | What it is | Use it when |
|---|
| Schedules | A time-based trigger on a bot | The job runs on a clock: daily checks, weekly reviews, monthly evidence |
| Listeners | An endpoint that receives events and webhooks | A tool reports something: a SIEM alert, a GuardDuty finding, a CI result |
| Service portal | Self-service apps your wider team can request | People need to ask for something, such as temporary access, and get it within policy |
| Optimus | The autonomous orchestrator, in a chat | You have an ad-hoc question or task, or want to build by describing the goal |
| API keys | Programmatic access | Another system should start or query autobotAI work |
Do the work
| Component | What it is | Use it when |
|---|
| Bots (workflows) | A sequence of nodes you can read step by step | The steps are known and should run the same way every time |
| Actions | A reusable operation in an integration: library, no-code or Python | A workflow or agent needs to read from or change a system |
| Action types | CLI, Terraform, Ansible, kubectl, Bash, SSH, Steampipe | The fix or check is infrastructure work |
| AI evaluator node | An AI step that classifies, scores or decides | A rule can't express the judgment, such as "is this alert real?" |
| AI agent node | An agent with tools and a goal at one step | One step needs open-ended reasoning, inside an otherwise fixed workflow |
| Logic nodes | Condition, loop, merge, delay, parameters | You need branches, repeats, waits or reusable values |
| Harnesses | Agents, triggers, memory and approvals assembled around one goal | The outcome spans many runs, tools and owners. See Choosing a workflow or a harness |
Give agents context
| Component | What it is | Use it when |
|---|
| Memory spaces | Searchable stores that agents read from and add to | Agents need to recall earlier findings, asset relationships or owners across runs |
| Knowledge | Policies, runbooks and reference material for Optimus | Answers should follow your rules, not general advice |
| Skills | Versioned, scanned packages of know-how | Several agents should do a job the same way, as your team does it |
| Prompt management | How prompts are written and governed | You are tuning AI nodes and agents |
Keep people in charge
| Component | What it is | Use it when |
|---|
| Approval node and Approvals | A pause until a named person decides | A step changes something that matters |
| Input node (User Inputs) | A form that asks a person for information | A workflow needs details only a person has |
| Notifications | Messages to Slack, Microsoft Teams, Google Chat or email | Owners and stakeholders need to know what happened |
| Autonomy tiers | A0 advisory to A3 constrained by design | You decide how much each action may do on its own |
| Role-based access | Who can build, run, approve and administer | Several teams share a workspace |
Track and prove
| Component | What it is | Use it when |
|---|
| Bot executions | Every run, with each step's input and outcome | You need to see what a workflow did, or why it failed |
| Guardian tasks | The queue of findings agents are working, with status and owner | You want one view of open work across tools |
| Workloads | Applications and services that findings are recorded under | Findings should roll up to the system and team that owns them |
| Risk register | Accepted risks with owner, approver, justification and expiry | A finding won't be fixed now and the exception must be on record |
| Dashboards and Compliance | Metrics, posture and framework views | Leaders and auditors need the picture without reading runs |
| Audit trails | Who did what, when and why | Evidence for audits and investigations |
Need a hand with this?Our engineers can walk through it with you on your own stack.
