Reference
AI Context Reference
How OpsCompanion's AI agent reasons about your infrastructure and what actions it can take.
OpsCompanion provides a context-aware AI agent that understands your specific infrastructure, not generic answers from the internet.
This page describes what the AI agent knows, the actions it can take, and how operational memory shapes its responses.
What Makes It Context-Aware
OpsCompanion's AI agent is not a generic chatbot. It reasons over a continuously updated operational context layer built from your stack:
- Your specific infrastructure resources and configuration
- Dependencies and relationships between services
- Change history: what changed, when, and by whom
- Business context and notes your team has added
- Operational memory accumulated across past investigations and incidents
This context makes answers specific to your infrastructure and grounded in observed data, not speculation.
Agent Actions
The AI agent can perform a range of actions across your stack. The actions available depend on your plan.
Available on All Plans
| Action | Description |
|---|---|
| Investigate & summarize an incident | Trace what happened across services, correlate changes, and produce a summary |
| Answer "What changed?" questions | Surface recent changes across infrastructure, code, and deployments |
| Log to code correlation | Connect log entries back to the code and commits that produced them |
| Deploy context surfacing | Show what was deployed, when, and what it affects |
Starter and Above
| Action | Description |
|---|---|
| Pattern detection | Identify recurring issues and trends across your stack over time |
| Cost analysis & visibility | Track cloud spend, catch anomalies early, identify optimization opportunities |
| Documentation updates from changes | Keep operational context current as infrastructure evolves |
| Git + AI session capture indexing | Track changes made by AI coding tools alongside human commits |
| Manual change brief generation | Generate a summary of what changed across your stack on demand |
Growth and Above
| Action | Description |
|---|---|
| Scheduled daily commit brief | Automated daily summary of commits and changes across your stack |
| Cross-workspace reasoning | Query and correlate across multiple environments and workspaces |
| Workflow execution | Execute multi-step operational workflows |
| Draft pull request | Create pull requests with code changes from investigation context |
Enterprise Only
| Action | Description |
|---|---|
| Guardrailed infra changes | Make infrastructure changes with policy enforcement and approval flows |
| Autonomous remediation | Automated response to known failure patterns with guardrails |
Operational Memory
The AI agent's answers improve over time because OpsCompanion builds persistent operational memory.
What Memory Captures
- Infrastructure state and configuration changes
- Incident investigations and their outcomes
- Team-added context, notes, and ownership information
- Commit history and AI agent session activity
- Patterns detected across deployments and services
Memory Capabilities by Plan
| Capability | Free | Starter | Growth | Enterprise |
|---|---|---|---|---|
| Raw Memory Stored | Unlimited | Unlimited | Unlimited | Unlimited |
| Searchable History | 7 days | 30 days | 90 days | Full history |
| Long-Term Pattern Detection | Not available | 30-day window | 90-day window | Full history |
| Historical Incident Recall | Limited | Recent | Extended | Full archive |
| Cross-Workspace Memory Queries | Not available | Not available | Available | Available |
| Multi-Agent Memory Fusion | Not available | Not available | Available | Available |
| Persistent Organizational Memory | Not available | Not available | Available | Available |
How Memory Improves Responses
When you investigate an incident, the AI agent draws on memory to:
- Recall similar past incidents and how they were resolved
- Surface changes that correlate with the current issue
- Identify patterns the team may not have noticed
- Provide context that would otherwise require asking someone who was there
Memory means your team's operational understanding compounds over time instead of resetting with each incident.
What the AI Knows
Infrastructure Context
- Resource inventory across all connected cloud providers
- Configuration and metadata for each resource
- Dependency relationships derived from observed data
- Change history from audit logs and provider APIs
Business Context
- Manual links your team has created between resources
- Notes explaining business purpose and ownership
- Relationships that only humans know about
- Context captured from past investigations
What You Can Ask
Questions the AI can answer using your specific context:
- What resources exist in this environment?
- What depends on this service?
- What changed in the last 24 hours?
- Who owns this resource?
- What depends on this and what breaks if I change it?
- What resources are related to this incident?
- How can I make this more cost effective?
- Am I using this resource to the best of my ability?
- Are there services I'm paying for that aren't being used?
- Has this kind of failure happened before?
How Answers Are Grounded
AI responses are constrained by your operational context:
- Answers reference actual observed infrastructure state
- Relationships come from real configuration data
- Change information comes from provider APIs and audit logs
- Business context comes from notes and links your team has added
- Memory provides historical patterns and past investigation outcomes
This reduces speculation compared to generic AI systems.
Current Constraints
- OpsCompanion AI currently queries and does not execute infrastructure changes (except on Enterprise with guardrails)
- Answers are limited to what the connected integrations provide
- The AI does not replace monitoring or alerting systems
Supported Data Sources
The AI reasons over data from:
- AWS - EC2, Lambda, S3, RDS, DynamoDB, CloudWatch, CloudTrail, IAM, and more
- GCP - VMs, Cloud Run, Cloud SQL, Cloud Storage, Pub/Sub, and more
- Azure - VMs, SQL databases, storage accounts, AKS, security assessments, and more
- DigitalOcean - Droplets, Kubernetes, databases, Spaces, and more
- GitHub - Repositories, PRs, issues, commits, deployments
- Vercel - Projects, deployments, domains
- Lovable - AI-generated web applications
- Base44 - AI-generated web applications
- Business context - Links, notes, and ownership your team has added
- Operational memory - Past investigations, incident history, and detected patterns
For a full breakdown of per-provider capabilities, see the AI Capabilities reference.