
SkaleAgents
An IaC and architecture review MCP service with a built-in AI assistant

Tool details
SkaleAgents is a "bring your own AI" IaC and architecture review service for DevOps: it neither trains nor hosts any model itself, it only runs 48 deterministic rules and then hands the results over an MCP Server to the Claude, Cursor, or ChatGPT you already use for explanation. The README puts it plainly: No separate hosted model is invoked.
Core capabilities
- MCP tool calls: four tools such as review_architecture and scan_iac plug into your editor, so you can start a review from your own AI assistant
- 48 deterministic rules: 37 security, 6 reliability, 3 AI workload, and 2 scaling and cost, returning rule IDs and remediation advice
- 18 expert bots: the site's own DevSecOps, SRE, and MLOps prompts, free to copy into any model
- Scheduled audits and Watchtower: set cron and time zone per plan, push results over webhook, collaborate in shared sessions
- CLI and GitHub Action: npx @skaleagents/swarm scan supports --changed and --fail-on high, and runs offline without reporting back
- Secret scanning: SECRET001 covers AWS, GitHub, Stripe and other keys plus private keys, reporting the location only and never echoing plaintext
Who it fits Small teams of 5 to 50 people without a dedicated DevSecOps engineer want one look at whether their Terraform or Kubernetes is left exposed before merging; developers already using Cursor would rather call the rules straight from the conversation.
What sets it apart The open-source scanners Checkov and tfsec only hand you rules, K8sGPT needs a real cluster, and commercial AIOps platforms charge a standing monthly fee. It takes the middle path: it does not connect to your cloud, it bills per run, and it hands deterministic results to your own model to explain, starting at $19.
Editor's take A good fit for teams that want to make a "pre-release checkup" routine but still do not want to connect a cloud account; if you need real-time inspection in production or compliance certification, it explicitly does not do that, and you should look to K8sGPT or Komodor instead. The Free plan gives 2 reviews per month, and the rule documentation is fully open.
Pricing
Freemium$0 (Free plan: 2 audit runs per month)2 audit runs per month, good for trying out the rule coverage
25 reviews per month, 1 scheduled audit, minimum interval 48 hours
200 reviews per month, 10 scheduled audits, minimum interval 24 hours
2,000 reviews per month, 100 scheduled audits, minimum interval 1 hour
Custom quotas and support negotiated by team size
Additional notes
Billed monthly, with no published annual price; quotas are counted in audit runs, and the Free tier is a long-term plan rather than a one-time trial
FAQ
User reviews
User Reviews
No reviews yet
No one has reviewed this tool yet