Bot directory01
Find your specialist
Search by task and filter by specialty. Read a bot's instructions, then open them in your assistant with your own project context.
Explore botsBuilt for the work behind the deploy
SkaleAgents
Connect the AI you already use.
It checks your application and your infrastructure, then tells you what to change.
Public bots and prompts. No account needed to browse.
Built on MCP / Connected to your workflow
Your editor is the starting point.
Connect the AI you already use. It checks your application and your infrastructure, then tells you what to change. The findings come back in the same conversation.
Set up your MCP connectionCursor · Claude Code · Claude Desktop · ChatGPT · Codex
01 / Configure
Add SkaleAgents to your client
Choose your client on the MCP page and add the hosted server URL.
02 / Authorize
Connect through Google OAuth
Connect in your client, sign in if needed, and approve MCP access. Credentials are handled automatically.
03 / Request
Ask for an architecture review
Have your assistant call
review_architecturewith relevant source or infrastructure. Read the findings in the same conversation.
01 / Put it to work
What would you like
a second pair of eyes on?
A useful review connects a finding to the code behind it. Pick a focus to see the input, the concern, and a practical next step.
Auto preview / 9 seconds
Choose a review
DevSecOps / review brief
Illustrative exampleCatch the open door before you ship.
Bring a Terraform snippet or a Kubernetes manifest. Ask for exposed services, broad permissions, and the smallest useful fix.
01 / What you bring
security-group.tf
ingress {
from_port = 22
to_port = 22
protocol = "tcp"
cidr_blocks = ["0.0.0.0/0"]
}
resource "aws_db_instance" "app" {
publicly_accessible = true
}02 / What a review looks like
SSH is reachable from any IPv4 address.
Port 22 accepts traffic from 0.0.0.0/0. Any host on the internet can attempt a connection if the instance is publicly reachable.
Public database
publicly_accessible = true becomes publicly_accessible = false
Suggested next step
Restrict ingress to your approved access network, or use a session-based access method. Confirm the access path before removing the rule.
Try this approach with your own context. These examples are not live audit results.
Read or customize the prompt
Try it yourself / No account needed
Change a setting.
See why it matters.
Edit the simplified deployment settings below. Three local rules update the sample findings as you type. This is an interactive example, not a live AI audit or a deployable manifest.
Runs in your browser
Try replicas: 2, turn readinessProbe on, or set memoryLimitMi to a positive number.
What does this demo check?
- Whether at least two replicas are configured.
- Whether the readiness-probe setting is enabled.
- Whether a positive memory limit is supplied in Mi.
The example improvements illustrate these rules. Real values depend on your workload. No code is executed or sent to an API.
Sample findings
0 of 3 sample checks configured
replicasReview thisA single replica
If that instance stops, there is no second replica to serve requests. Consider additional replicas for availability.
readinessProbeReview thisReadiness checks are off
Enable a readiness probe so traffic waits until the application is ready to accept it.
memoryLimitMiReview thisNo memory limit set
Choose a memory limit using workload measurements so one container cannot consume memory without a configured bound.
02 / Your tools, your workflow
Start where you already work.
Use a prompt in chat, connect a tool in your editor, or review an audit with your team. Choose the setup that fits the task.
Start with a specialist, not a blank chat.
A bot packages a specialist's instructions into a reusable prompt. You bring the project context and run it in your preferred assistant.
Find a bot
Browse the public directory by specialty and open a bot to read its instructions.
Choose your assistant
Open Actions and send the prompt to Claude, ChatGPT, Grok, Codex, or Perplexity. Copy it if you use Gemini or another agent.
Add your context
Provide relevant code or configuration. Ask for evidence and validation steps before applying a recommendation.
Keep the review beside your code.
MCP (Model Context Protocol) lets a compatible AI client call SkaleAgents tools. Add https://skaleagents.com/mcp and approve OAuth.
Add the hosted server
Copy https://skaleagents.com/mcp and choose OAuth in your client.
Sign in with Google
Click Allow to approve MCP access. Return to your editor to continue.
Ask for a review
Have your assistant call review_architecture with relevant source or IaC and a security, reliability, cost, or general focus.
Take a finding into a shared review.
Submit infrastructure text as an audit job, then inspect the findings in a Watchtower session with your team.
Scope an audit
Sign in, paste your IaC, and select DevSecOps, SRE, MLOps, or scaling specialists. Add telemetry JSON when it helps.
Review the results
Open Watchtower from the job to inspect findings, share the session, and send a redirect instruction.
Repeat when needed
Create a scheduled audit with a cron expression and timezone. An optional webhook receives completion or failure alerts.
03 / Inside SkaleAgents
From the first question
to a repeatable review.
Creator Studio02
Write a review playbook
Start with a security, reliability, or capacity template. Customize the instructions, save a draft, and publish publicly or keep it private.
Open StudioPrompt library03
Find it. Copy it. Get to work.
The library lists title, type, and category. Open a prompt to read it and copy the full instructions. Sign in to save your own reusable prompts.
Browse promptsPublic reference04
Read the review checks
See what each built-in check looks for, the input it needs, and what a result does not prove. No account required.
Read the checksScheduled audits05
Set a review rhythm
Review saved infrastructure text or a public GitHub file on a schedule. Set the timezone and send audit alerts to a webhook.
View schedulesMCP connection06
Stay in your editor
Connect the AI you already use. It checks your application and your infrastructure, then tells you what to change.
MCP config04 / Before you start
Good questions.
Straight answers.
What you can use, what you share, and where the boundaries are.
Looking for connection details?What is SkaleAgents, in plain English?
SkaleAgents gives developers reusable specialist prompts and tools for reviewing code and infrastructure. You can take a bot into an AI assistant, call review tools from your editor with MCP, or submit an audit in the web app. It is designed for engineers and small teams who want a more structured DevOps review.
How is this different from asking an AI chatbot?
You can still use your own AI assistant. SkaleAgents adds reusable specialist instructions, a prompt library, and a tool connection for architecture reviews. The web workflow keeps audit jobs and shared Watchtower reviews in one place. A bot is a set of instructions, not a separately trained model.
Can I try it without an account?
Yes. Browse the public bot directory, community prompts, free resources, the blog, and review checks without signing in. Google sign-in is needed to create or manage your own bots and prompts, authorize an MCP connection, and use account-owned audit workflows.
What does it cost?
Public bots, community prompts, and the review checks are free to read. Organization audit and monitoring plans are on our pricing page. An owner chooses Starter, Pro, or Max for an organization they manage. Using a prompt in another AI assistant is subject to that provider's pricing and usage limits; SkaleAgents does not include a subscription to those services.
Which AI assistants and editors can I use?
Prompt and bot pages can open Claude, ChatGPT, Grok, Grokbot, Codex, or Perplexity. Gemini and Grokbot copy the text first because those destinations do not take a prefilled prompt. You can also copy plain text, markdown, or a handoff written for another agent. For tool calls, connect Cursor, Claude Code, Claude Desktop, ChatGPT, or Codex to the hosted MCP server through OAuth. See MCP config for each client's setup and plan requirements. Opening a prompt does not link that assistant to your SkaleAgents account.
Do I need to connect my cloud account or repository?
Not for the text-based workflows shown here. You supply code, infrastructure text, and optional telemetry. The MCP client sends the content included in its tool calls; the web audit uses the text you submit. Review what you share and remove credentials before sending it. A review can only assess the context it receives.
Will an agent change my production infrastructure?
These workflows produce findings and recommendations for you to review. Submitting an audit or opening a Watchtower session does not itself deploy a change. Validate proposed fixes and apply them through your own development and deployment process.
Can I keep my bots private and revoke tool access?
Yes. Creator Studio supports public and private bots, with drafts kept out of the public directory. MCP clients are authorized through OAuth and tied to your Google-backed account. Disconnect a client from MCP connections in your dashboard to remove its access.
What are the current limits?
Reviews depend on the supplied code and context; they are not a guarantee that an application is secure or compliant. Schedules can check one file in a public GitHub repository; private repositories and cloud accounts are not connected. Cloudploy deployment handoff, Crontinel telemetry integration, and enterprise compliance certification are not promised by the workflows on this page.
Your next step
Find your starting point.
Two choices. A suggested workflow. Explore the options here before opening the app.
