Runs locally on your workstation
Oscar executes entirely on your machine. Your cloud data is queried locally and never transmitted to OpsCanvas infrastructure. What happens in your environment stays in your environment.
Free Operator Edition · No credentials shared
Oscar Ops
Your AI co-pilot for daily cloud operations.
AI coding agents are enabling code to ship faster than ever. Oscar helps your operations keep up.
Oscar Ops lives in your CLI. It connects to your cloud accounts, deployment tools, observability stack, and issue trackers, builds a live context graph of your environment, and gives you cross-system answers in plain English in seconds, not hours. No new dashboard. No credentials shared. No changes made without your approval.
Runs on your workstation
Bring your own AI model
Works on untagged environments
No action without your approval
<30min
From install to your first context-grounded cloud answer
10+
AI models supported, use the one you already have
20+
Integrations across cloud, CI/CD, observability, and issue tracking
50%
Fewer tokens per query vs unstructured LLM prompting
0
Cloud credentials or data ever shared with OpsCanvas
Why Not Just Use Claude or ChatGPT?
Purpose-built for cloud operations.
Oscar is different. Questions are answered with live context, persistent memory, team coordination, and workflow integration that generic LLMs are not made to do well.
| Capability | What Oscar Ops does |
|---|---|
| Environment knowledge | Live context graph of your accounts, resources, dependencies, and agents, loaded automatically before you type. |
| Deep contextual analysis | Contextual reasoning over your actual topology. Oscar can answer "why is this service slow" with real dependency data, not generic advice. |
| Team coordination | Shared context graph across the team. Oscar Pro enables task allocation, handoffs with full context, role-based oversight, and guardrail enforcement for everyone. |
| Workflow integration | Designed for complex operational workflows. Oscar can receive alerts, investigate, propose fixes, and route approvals as part of a larger automated system. |
| Persistent memory | Remembers your environment across sessions. Picks up where you left off without re-pasting context. |
| Action safety | Approval required before any action executes. Oscar knows what each change can affect and will not propose changes that exceed your permissions. |
| Token efficiency | Structured context from the graph means 50% fewer tokens per query, more accurate answers, and lower cost. |
Context is the Difference
Context is the difference.
An agent powered by a live context graph outperforms all others, every time.
Incident response
Root cause analysis
Prompt
"payment-service is throwing 5xx errors. What happened in the last hour?"
Why context is critical
Root cause requires correlating a CI/CD deploy, a secrets config change, an EC2 autoscaling event, and a LaunchDarkly flag flip across four identity systems with no shared schema. The context graph already did that work.
Proactive operations
"What should I care about?"
Prompt
"It's quiet. What should I be worrying about that I'm not?"
Why context is critical
A generic LLM can answer questions you ask. It cannot tell you what questions to be asking. Oscar has history, pattern recognition, and memory of what was quietly concerning last week.
Access hygiene
Permission audit
Prompt
"Who can actually delete data from the customer-data S3 bucket?"
Why context is critical
Effective AWS permissions require graph traversal across roles, policies, SCPs, and permission boundaries. Oscar handles it and returns plain-language results. Generic models produce confident-sounding answers that are wrong in subtle ways.
Daily operations
Morning health check
Prompt
"What needs my attention this morning?"
Why context is critical
A 4% error rate is normal for one service and a five-alarm fire for another. Oscar knows your baselines from 90 days of history. It flags what is abnormal for your environment, not what is loud in the abstract.
Learning loops
Incident post-mortem
Prompt
"Draft a post-mortem for last night's checkout-service incident."
Why context is critical
Reconstructing a timeline from raw cloud data is the most tedious part of post-mortem writing. Oscar walks the context graph with the incident time range as a filter and produces a complete draft in seconds. Your time goes to analysis, not archaeology.
Security Standards
Built for your security standards.
Oscar was designed for platform engineers who are accountable for what runs in production. Every design decision starts from the assumption that security is non-negotiable. Your data stays in your environment. Always.
Uses your existing credentials
Oscar uses the same credentials and access rights you already have. No new IAM roles to create, no service accounts to maintain, no credential handover to a third party.
Your AI model, your account
Oscar works with the foundation model of your choice. Your model API key stays on your machine and is billed directly to your account. OpsCanvas never sees your prompts or model credentials.
Human approval before every action
Oscar analyzes and proposes. It never executes a change without explicit human sign-off. Every proposed action is shown with its full evidence trail before you approve or reject it.
Full audit trail
Every query, finding, proposed action, and approval decision is logged with its full decision trace, ready for compliance review, post-mortems, or governance audits without manual reconstruction.
Permission-aware reasoning
Oscar only suggests actions within the scope of your existing access rights. It understands what you are allowed to do and will not propose changes that exceed your permissions.
Get to the Why
Get to the why in minutes.
Every cloud investigation starts the same way: what changed, why did it fail, who owns it, what depends on it, and is it safe to act. Those answers exist scattered across a dozen tools. Oscar connects them and gives you one answer.
Incident investigation and root cause
Oscar helps you investigate incidents and operational questions across your environment in seconds instead of hours. By pulling signals from cloud accounts, deployments, clusters, logs, and resources simultaneously, Oscar correlates the data to explain what happened and why, and surfaces the evidence you need for a full post-mortem.
✓
Correlates CloudTrail events, deployment history, config changes, and cost in one pass ✓
Eliminates the 20-minute archaeology session at the start of every incident ✓
Generates structured incident summaries ready to paste into tickets or post-mortems ✓
Persistent memory: picks up where you left off, no re-pasting context Incident: api-gateway latency spike
Investigating
Root cause identified
2m 14s
Deploy v1.9.2 at 02:14 UTC introduced missing Lambda concurrency limit. ECS autoscaled to 34 instances. CloudTrail confirms: james@eng via GitHub Actions pipeline #4821.
Blast radius
3 services
payments-api, auth-service, reporting-cron all depend on api-gateway. No data loss detected. SLA breach window: 4m 32s.
Proposed fix
Ready to apply
Set Lambda reserved concurrency to 200. Estimated savings: $1,200/mo. Awaiting approval from ops-lead@eng.
Cost analysis and spike investigation
Oscar analyzes cloud spend across accounts and services, surfaces cost trends, anomalies, and unexpected spikes in seconds. When something stands out, Oscar flags it and lets you drill into the details, connecting the spend back to the deployment, the team, and the specific resource that caused it.
✓
Attributes cost to the specific resource, team, IaC module, and deploy event that caused it ✓
Surfaces cost anomalies with root cause, not just a spike on a chart ✓
Identifies idle and over-provisioned resources with safe-to-remove evidence ✓
Generates CFO-ready summaries you can paste directly into Slack or a ticket $ oscar /cost --spike --since 7d
⚠ Spike detected: +$8,400 vs prior 7-day period
ECS cluster: api-gateway · autoscaled Tue 02:14 UTC
Triggered by: deploy/api-gateway v1.9.2 · james@eng
Root cause: missing Lambda concurrency limit
Estimated fix savings: $1,200/mo · Fix ready to propose
$ oscar /act --apply-fix --request-approval
✓ Approval request sent → james@eng, ops-lead@eng
No changes made. Waiting for sign-off.
Cloud health in seconds
Oscar scans your cloud accounts, Kubernetes clusters, and connected systems and gives you a clear view of what is healthy, what needs attention, and where issues are flagged across every account in one pass. For larger environments, Oscar generates a local report you can review visually before continuing your investigation.
✓
Multi-account, multi-region scan in a single command ✓
Surfaces flagged resources with full context: who deployed it, when, and what depends on it ✓
Understands your environment's baselines and flags what's abnormal, not just what's loud ✓
Works on untagged environments from the first scan oscar /check --all
Live
us-east-1 · 14 accounts
Healthy: 312
No critical issues detected across primary accounts.
ELB: backend-lb-prod-08a9b2
Attention
Deployed by: sarah@eng · PR #2847 · 6 months ago · Project: backend-v2 (decommissioned) · 0 active connections · Cost: $40/mo
AI agent: infra-automation-prod
No Policy
Access: EC2 full, S3 full · Owner: unknown · Token cost: $1,240/mo · No approval workflow
Ask your cloud anything
Use natural language to ask questions about trends, anomalies, or system behavior across your entire operational stack. Unlike point solutions limited to a single system, Oscar connects across your cloud accounts, CI/CD pipeline, observability tools, and issue trackers to deliver complete answers with the context cloud engineers actually need.
✓
Answers questions that span multiple tools and accounts in a single response ✓
No copy-pasting logs into a chat window. Oscar queries the source directly ✓
Shareable summaries formatted for Slack, tickets, or email in one command ✓
If a needed integration is not listed, Oscar can learn to navigate new tools $ oscar "what changed in prod since Friday?"
Querying: CloudTrail · GitHub · Datadog · PagerDuty
✓ 4 changes detected across 3 services (Fri to Mon)
api-gateway v1.9.2 · james@eng · Fri 23:41 (caused spike)
auth-service v2.1.0 · priya@eng · Sat 14:22 (clean)
reporting-cron config · terraform · Sun 03:00 (clean)
ECS capacity rule updated · james@eng · Mon 08:55
1 PagerDuty alert · 0 open incidents · 2 Jira tickets created
How Oscar Fits
Not a replacement for your existing tools.
The layer that makes them work together.
Oscar is complementary to the tools you already run. It does not compete with Datadog, CloudHealth, or your ticketing system. It connects them, providing the cross-system context that each individual tool is missing on its own.
Datadog finds the problem. Oscar finds the person who can fix it.
Datadog sees telemetry.It tells you a service degraded and surfaces the relevant metrics and traces.
What Oscar adds:Oscar sees the full chain: code, cloud, IaC, tickets, ownership, and permissions. When Datadog fires an alert, Oscar tells you which deploy caused it, who owns the service, what else depends on it, and routes the fix request to the right person.
CloudHealth tells you spend went up. Oscar tells you why and routes the cleanup.
FinOps tools show you the bill after the fact.They allocate spend and generate reports.
What Oscar adds:Oscar surfaces cost anomalies in real time, attributes them to the specific resource, deploy event, and team that caused them, and proposes a governed fix. The answer goes from "your costs increased" to "here is the pull request, here is the owner, here is the proposed change."
Scripts automate known tasks. Oscar handles what changes every day.
Internal scripts and runbooks are brittle and static.They break when infrastructure changes and require constant maintenance.
What Oscar adds:Oscar is adaptive and remembers. It learns from every investigation and reasons about new situations using the current state of your environment, not a frozen script written six months ago.
A Week with Oscar
What your daily workflow actually looks like.
Oscar is not a feature you use once in a while. After setup, it becomes the first thing you reach for every morning and the fastest path through every incident. Here is what that looks like in practice.
Mon 9am
Morning Check-in
Start the week with a full picture in under 5 minutes
In 90 seconds Oscar scans all 14 of your AWS accounts, surfaces three items that need attention before standup, and flags an AI agent that appeared over the weekend with no owner and full EC2 access. You assign it to the platform team with one command before the meeting starts.
Tue 2am
Incident Response
On-call fires at 2am. You have root cause in 3 minutes.
PagerDuty wakes you up. You ask Oscar what happened. It correlates the alert with a deploy that went out 20 minutes earlier, identifies the missing concurrency limit, shows you the blast radius (3 downstream services), and proposes a fix for your approval. You approve it. You are back asleep by 2:18am. The post-mortem draft is ready before morning standup.
Wed 11am
Team Coordination (Oscar Pro)
Your teammate picks up exactly where you left off.
You have been investigating a cost spike all morning and need to hand it off for the afternoon. With Oscar Pro, you assign the investigation to a teammate and they receive full context: every query you ran, every finding Oscar surfaced, and the proposed fix still pending approval. They pick it up in 2 minutes with no Slack thread archaeology.
Fri 4pm
Proactive Hygiene
Ask what you should be worried about before the weekend.
You ask Oscar what looks concerning heading into the weekend. It surfaces a backup configuration that never ran successfully, a zombie ELB that has been accumulating cost for six months, and a service with an RTO target of 4 hours but no backup policy. You create tickets for all three and head into the weekend with visibility you would never have had from manual checks.
Supported Models
Use the model you already have.
Oscar works with all major foundation models. Your API key stays on your machine. You are billed directly by your model provider. OpsCanvas never sees your prompts or model credentials.
Anthropic Claude
OpenAI GPT
Meta Llama
Mistral
DeepSeek
Google Gemini
xAI Grok
Qwen
Perplexity
Cohere
Integrations
Works with the tools you already run.
Oscar connects to your existing stack at setup. No new dashboards, no migration, no rip-and-replace. If an integration you need is not listed, Oscar can learn to navigate new tools.
AWS
Azure
Google Cloud
GitHub
GitLab
Bitbucket
Datadog
New Relic
Dynatrace
Grafana
CloudWatch
Azure Monitor
OpenTelemetry
Jira
Linear
Honeycomb
App Insights
Get Started
Install Oscar. Get answers in 30 minutes.
Four steps from download to your first context-grounded cloud investigation. No new accounts, no onboarding calls, no dashboard to configure.
Scale with Your Team
Start free. Grow when you're ready.
Oscar Ops Operator Edition is free forever for individual engineers. When you are ready to share context across your team or connect assessments and workflows to the platform, we have you covered.
Oscar Pro
For Engineering Teams
Share investigations, collaborate on incidents, and build a shared context graph that every engineer on your team can query. When someone handles the 3am incident, the next engineer picks up full context, not a Slack thread summary.
$00/ seat / month
- Shared context graph across the team
- Collaborative incident investigation with full handoff context
- Task allocation and assignment across engineers
- Role-based controls: who can query, propose, or act
- Team-level audit trail for governance reporting
- Shared skills library and guardrail enforcement across the team
OpsCanvas Platform
Assessments & Workflows
When you are ready to go beyond daily operations, OpsCanvas delivers structured assessments of your backup posture, AI agent inventory, and cloud waste, plus platform workflows that close the gaps at scale.
Talk to us about your needs
- Backup Assessment: evidence-backed posture report in days
- AI Agent Inventory: complete governance picture of every agent
- Zombie Waste Assessment: safe reclaim with full IaC lineage
- Ongoing monitoring to keep your posture current
- Platform workflows for remediation at scale
Common Questions
Everything you need to know
before you install.
Why not just use Claude or ChatGPT for my cloud questions?
Generic LLMs answer questions about cloud concepts. Oscar answers questions about your cloud. The difference is the context graph: a persistent, automatically maintained model of your specific environment including what runs, what owns what, what depends on what, and what has changed. You cannot paste that into a chat window. It takes time to build, and Oscar builds it automatically. Beyond context, Oscar is designed for team coordination (shared investigations, task allocation, role-based guardrails), deep workflow integration (it can fit inside multi-step operational pipelines, not just respond to typed questions), and safety (it knows your blast radius before proposing any action). Oscar also uses structured context from the graph, which means roughly 50% fewer tokens per query compared to manually pasted prompts.
Is Oscar secure? Will it create problems in my environment?
Oscar runs entirely on your workstation. Your cloud data is queried locally and never transmitted to OpsCanvas infrastructure. Oscar uses your existing credentials and access rights, so it operates within exactly the permission boundary you already have. Your AI model API key stays on your machine, billed directly to your model provider. Oscar never executes any change without explicit human approval: every proposed action is shown with a full evidence trail first. There is nothing to approve at the infrastructure level beyond granting Oscar the same access you already have as an engineer.
How does Oscar build context without me doing anything?
When you run oscar /setup, Oscar scans your accessible cloud accounts, connected tools, and IaC repositories using your existing credentials. It infers ownership and dependencies from runtime behavior, not from tags. That means it works on untagged environments from day one. The context graph grows over time as you run more commands: each investigation, health check, and cost query adds to the model. By the end of your first week, Oscar has a richer picture of your environment than most tagging-based systems ever achieve.
Which AI models does Oscar support?
Oscar supports all major foundation models including Anthropic Claude, OpenAI GPT, Meta Llama, Google Gemini, Mistral, DeepSeek, xAI Grok, Qwen, Perplexity, and Cohere. You configure the model of your choice at setup. Your API key stays on your machine and you are billed directly by the model provider. OpsCanvas never sees your prompts or model credentials. You can switch models at any time without losing your context graph.
What does Oscar Pro add for teams?
Oscar Operator Edition is a powerful individual productivity tool. Oscar Pro extends it into team infrastructure. With Pro, your team shares a single context graph: when one engineer investigates an incident, their findings are immediately available to teammates. You can assign tasks, hand off investigations with full context (not a Slack summary), enforce role-based guardrails so different team members have different levels of authority to propose or execute actions, and see a team-level audit trail for governance reporting. Pro also includes shared skills and weekly operational digests for the team.
How is Oscar different from point solutions like Datadog AI or CloudHealth?
Oscar is complementary to those tools, not a replacement. Datadog sees telemetry and tells you something degraded. Oscar sees the full chain: code, cloud, IaC, tickets, ownership, and permissions, and tells you who deployed the change, why costs went up, what else breaks if you act, and who should approve the fix. CloudHealth shows you spend after the fact. Oscar surfaces anomalies in real time and attributes them to specific resources, teams, and deploy events. Point solutions are deep in one domain. Oscar connects them all so you can get to the answer without opening six browser tabs.
Download Oscar Ops Today
Your cloud context graph.
Up and running in 30 minutes.
Free to install. Works with the tools you already use. No new dashboard. No credentials shared. Your first context-grounded cloud answer in under 30 minutes.
Free Operator Edition · Bring your own AI model · Works with untagged environments · No credentials shared