Documentation

DevOps Agent
Docs

Everything you need to connect your cloud infrastructure, automate deployments, and manage incidents โ€” all through a natural language chat interface.

๐Ÿ’ฌ
Chat with your infrastructure
Ask questions, run commands, and get insights in plain English โ€” no CLI required.
โ˜๏ธ
727 cloud tools
AWS, Azure, GCP, and Kubernetes โ€” full lifecycle management from a single interface.
๐Ÿ”
Safe by default
Destructive operations require explicit confirmation. All credentials are encrypted at rest.
๐Ÿ“š
Learn as you go
Learning Mode explains every action inline so you build knowledge while getting work done.

Getting Started

You can be up and running in under five minutes. Here's the full flow from registration to your first cloud command.

1
Create your account
Go to /signup and register with your email address. You'll receive a verification email โ€” click the link to activate your account.
2
Complete onboarding
After email verification you'll be taken through a short onboarding flow. Choose your experience level (Beginner / Intermediate / Advanced) and the cloud providers you use.
3
Connect a cloud provider
Navigate to Settings โ†’ Cloud Credentials and add your AWS, Azure, or GCP credentials. See the Cloud Integrations section below for detailed setup instructions per provider.
4
Send your first message
Head to the Chat tab and type something like list my EC2 instances in us-east-1. The agent will call the appropriate tool and return results in plain English.
โ„น๏ธ Free tier: New accounts start with 20 credits/month on the Free plan, giving you full access to the agent with real tool execution. Upgrade any time from the Billing page.

Chat Interface

The chat interface is the primary way to interact with DevOps Agent. You talk to the AI in plain English and it figures out which tools to call, in what order, and how to present the results.

Example commands

# EC2 management
List all running EC2 instances in us-east-1
Stop instance i-0abc123def456789
Create a t3.micro instance named web-server-01

# Cost & optimization
What are my top 5 most expensive AWS services this month?
Show me idle EC2 instances I can terminate

# Kubernetes
Show pods in the production namespace
Scale the api-deployment to 4 replicas
Get logs for pod api-server-xyz

# Incident response
Run the high-cpu-usage playbook on app-server-02
Show me all firing alerts in the last 2 hours

Confirmation for risky operations

Before executing any destructive action (deleting resources, terminating instances, modifying IAM policies), the agent will pause and ask you to confirm. This cannot be disabled โ€” it is a core safety feature.

Model selector

Use the model toggle in the chat toolbar to switch between AI providers:

ModelProviderBest for
Kimi K2.5Moonshot AIFast responses, free/starter tiers
Claude SonnetAnthropicComplex multi-step reasoning (Professional)
Claude OpusAnthropicDeepest analysis, largest context (Business)

Credits

Each message deducts one credit. Your remaining credits are shown in the chat header. Free-tier users receive 20 credits per month. Paid plans include a monthly credit allocation โ€” see the Billing section for details.

Dashboard

The Dashboard gives you a live overview of your cloud infrastructure without having to ask the agent. It auto-syncs every time you visit.

WidgetWhat it shows
Resource SummaryTotal EC2, RDS, S3, Lambda, and Kubernetes resources across all connected providers
Cost TrendsMonthly spend chart + month-over-month change
Cost ForecastProjected spend for the current month based on usage so far
RecommendationsCost savings and rightsizing suggestions from your last sync
Resource GroupsBreakdown by service (EC2, RDS, etc.) with expandable utilization detail
โ„น๏ธ Click Sync Now in the dashboard header to force a fresh pull from your cloud providers. Sync runs automatically when you first visit after a 10-minute gap.

Learning Mode

Enable Learning Mode in Settings โ†’ Preferences to have the agent explain every action it takes as it takes it. Ideal if you're new to a cloud provider or want to build knowledge while getting work done.

Experience levels

LevelWhat changes
BeginnerFull concept explanations, analogies, step-by-step reasoning, links to AWS/GCP/Azure docs
IntermediateConcise explanations, highlights tradeoffs and gotchas, skips basics
AdvancedMinimal commentary, focuses on results and edge cases only

Your proficiency per topic is tracked automatically. Check Settings โ†’ Learning โ†’ Proficiency Radar to see where you're strongest and where there are gaps.

Sandbox Mode

Sandbox Mode lets you experiment with any command without touching real cloud resources. All tool calls return simulated but realistic responses โ€” perfect for demos, onboarding, or testing destructive workflows safely.

โš ๏ธ When Sandbox Mode is active, a banner appears at the top of the chat. No real API calls are made to AWS, Azure, GCP, or Kubernetes. Toggle it off in Settings โ†’ Preferences when you're ready to work with live infrastructure.

Sandbox Mode simulates realistic failure scenarios too โ€” disk full, pod crash loops, service downtime โ€” so you can practice incident response without manufacturing real outages.

Amazon Web Services

DevOps Agent supports 155 AWS tools covering EC2, VPC, IAM, RDS, Lambda, ECS, EKS, S3, DynamoDB, CloudFormation, Route53, SNS/SQS, CloudWatch, ElastiCache, and more.

Connecting AWS โ€” Access Key mode

Best for individuals and small teams. Navigate to Settings โ†’ Cloud Credentials โ†’ AWS:

1
Create an IAM user
In the AWS Console โ†’ IAM โ†’ Users โ†’ Add User. Attach the ReadOnlyAccess policy to start, or a custom policy scoped to the services you want the agent to manage.
2
Generate access keys
Under the user โ†’ Security credentials โ†’ Create access key. Copy the Access Key ID and Secret Access Key.
3
Paste into Settings
Paste both values into Settings โ†’ Cloud Credentials โ†’ AWS โ†’ Access Key mode, set your default region, and click Save. Click Test Connection to verify.

Connecting AWS โ€” IAM Role mode (enterprise)

Best for enterprise setups and cross-account access. Instead of long-lived access keys, the agent assumes an IAM Role using STS. Under Settings โ†’ AWS โ†’ switch to Role ARN mode and provide:

FieldExample
Role ARNarn:aws:iam::123456789012:role/DevOpsAgentRole
External IDOptional โ€” use for cross-account trust policies
Regionus-east-1

Microsoft Azure

DevOps Agent supports 36 Azure tools covering Virtual Machines, Storage, SQL, NSGs, Container Instances, Disks, VNets, and Public IPs.

Connecting Azure

Navigate to Settings โ†’ Cloud Credentials โ†’ Azure and provide:

FieldWhere to find it
Subscription IDAzure Portal โ†’ Subscriptions
Tenant IDAzure Active Directory โ†’ Properties โ†’ Tenant ID
Client IDApp registrations โ†’ your app โ†’ Application (client) ID
Client SecretApp registrations โ†’ your app โ†’ Certificates & secrets โ†’ New client secret
Default Locatione.g. eastus
โ„น๏ธ Your service principal needs at least the Contributor role on the subscription for write operations, or Reader for read-only access.

Google Cloud Platform

DevOps Agent supports 40 GCP tools covering Compute Engine, Cloud Storage, Cloud SQL, GKE, Cloud Run, Pub/Sub, Secret Manager, BigQuery, Cloud Functions, and Cloud Logs.

Connecting GCP

1
Create a Service Account
GCP Console โ†’ IAM & Admin โ†’ Service Accounts โ†’ Create Service Account. Grant it the roles your use case needs (e.g. roles/compute.viewer, roles/storage.admin).
2
Download the JSON key
Service Account โ†’ Keys โ†’ Add Key โ†’ Create new key โ†’ JSON. A .json file will download.
3
Upload in Settings
Settings โ†’ Cloud Credentials โ†’ GCP โ†’ paste the entire JSON content into the Service Account Key field, enter your Project ID and region, click Save.

Kubernetes

DevOps Agent supports 28 Kubernetes tools covering pods, deployments, services, namespaces, ConfigMaps, secrets, ingress, nodes, and jobs.

Connecting a cluster

DevOps Agent uses your local kubeconfig file. Ensure the machine running the agent has a valid kubeconfig at ~/.kube/config (or the path set in KUBECONFIG) with credentials for the cluster you want to manage.

To use a specific context, include the context name in your message:

Show pods in the production namespace using context prod-cluster

Supported operations

ResourceOperations
PodsList, describe, logs, delete, exec
DeploymentsList, describe, create, update image, scale, delete
ServicesList, create, delete
NamespacesList, get, create, delete
NodesList, cordon, uncordon
ConfigMaps / SecretsCreate, delete
Ingress / JobsCreate, delete

CI/CD โ€” Jenkins & GitHub Actions

DevOps Agent supports 21 CI/CD tools โ€” 11 for Jenkins and 10 for GitHub Actions โ€” covering pipelines, builds, jobs, and workflow runs.

Jenkins (11 tools)

Set your Jenkins URL and credentials in Settings โ†’ CI/CD before using Jenkins tools. The agent uses the Jenkins REST API with Basic auth.

ToolDescription
list_jenkins_jobsList all jobs on the Jenkins server
get_jenkins_job_statusGet the current status of a job
get_jenkins_build_logsRetrieve console output for a specific build
trigger_jenkins_buildTrigger a job build (optionally with parameters)
stop_jenkins_buildAbort a running build
list_jenkins_pipelinesList all pipeline jobs
get_jenkins_pipeline_statusGet pipeline run status
create_jenkins_pipelineCreate a new pipeline job from a Jenkinsfile
delete_jenkins_jobDelete a job (high-risk, requires confirmation)
get_jenkins_queueView the build queue
get_jenkins_server_infoRetrieve server version and health info

GitHub Actions (10 tools)

Set your GitHub token (GITHUB_TOKEN) in Settings. The token needs actions:read and actions:write scopes.

ToolDescription
list_github_workflowsList all workflows in a repository
get_github_workflow_runsGet recent runs for a workflow
trigger_github_workflowTrigger a workflow dispatch event
cancel_github_workflow_runCancel a running workflow
get_github_workflow_run_logsDownload logs from a workflow run
list_github_actions_secretsList Actions secrets for a repo
create_github_actions_secretCreate or update an Actions secret
delete_github_actions_secretDelete an Actions secret
get_github_repo_infoGet repository metadata and stats
list_github_branchesList branches in a repository

Example prompts

Show me all failing Jenkins jobs Trigger a build for the deploy-prod pipeline with branch=main Get the last 50 lines of logs from build #142 of the api-service job Cancel the workflow run #9823 on the deploy.yml workflow in org/repo

Docker

DevOps Agent supports 22 Docker tools covering container lifecycle, image management, volumes, networks, and Docker Compose.

Requirements

Docker must be installed and running on the machine where the agent runs. The agent communicates with the Docker daemon via the Docker SDK for Python (docker package).

Supported operations

CategoryOperations
ContainersList, inspect, start, stop, restart, remove, exec, logs, stats
ImagesList, pull, build, remove, tag, push
VolumesList, create, remove
NetworksList, create, remove
ComposeUp, down, ps (via shell)

Example prompts

List all running containers Show the last 100 log lines from the nginx container Pull the latest postgres:16 image Remove all stopped containers Run docker compose up -d in /opt/myapp

Terraform

DevOps Agent supports 16 Terraform tools for managing infrastructure-as-code: workspaces, state, plans, and module operations.

How it works

Terraform tools operate on a working_dir โ€” the directory containing your .tf files. Terraform must be installed on the machine where the agent runs.

Supported operations

CategoryOperations
CoreInit, validate, plan, apply, destroy
StateList, show, move, remove resources
WorkspacesList, create, select, delete
Output & ProvidersShow outputs, list providers, format
Caution: terraform_apply and terraform_destroy are classified as high-risk operations. The agent will ask for confirmation and show the plan output before proceeding.

Example prompts

Run terraform plan in /infra/prod and show me what will change Apply the staging workspace in /infra/staging List all Terraform workspaces in my infrastructure directory Show the output values from the networking module

SonarQube

DevOps Agent supports 12 SonarQube tools for code quality analysis, quality gate checks, and issue management.

Configuration

Set SONAR_TOKEN and SONAR_BASE_URL (e.g., https://sonarqube.yourcompany.com) in your environment. The token needs at minimum Browse permission on projects you want to inspect.

Supported operations

CategoryOperations
ProjectsList projects, get project quality gate status and metrics
IssuesList issues by severity, type, or component; mark as resolved
Quality GatesList quality gates, get conditions
MetricsCoverage, duplication, technical debt, code smells, bugs, vulnerabilities

Example prompts

What's the quality gate status for the api-service project? Show me all blocker issues in the frontend project List all SonarQube projects and their coverage metrics Are there any critical security vulnerabilities in my codebase?

Nexus Repository

DevOps Agent supports 12 Nexus Repository Manager tools for artifact management, repository administration, and component search.

Configuration

Set NEXUS_BASE_URL, NEXUS_USERNAME, and NEXUS_PASSWORD in your environment. The agent uses the Nexus REST API v1 with Basic authentication.

Supported operations

CategoryOperations
RepositoriesList, get details, create hosted/proxy, delete
ComponentsSearch, list, get, delete artifacts
AssetsList assets in a repository
SystemHealth check, node info, tasks list

Example prompts

List all repositories in Nexus Search for all versions of com.example:api-service in the maven-releases repo Delete old SNAPSHOT artifacts older than 30 days from the snapshots repository Check the health status of our Nexus instance

Ansible

DevOps Agent supports 13 Ansible tools for playbook execution, inventory management, and role discovery.

Requirements

Ansible must be installed on the machine where the agent runs. Playbook tools operate on a working_dir containing your playbooks and inventory.

Supported operations

CategoryOperations
PlaybooksRun playbook, run with tags, run with extra vars, dry-run (check mode)
InventoryList hosts, list groups, ping all hosts
Roles & GalaxyList installed roles, install from Galaxy
Ad-hocRun ad-hoc module commands against host patterns

Example prompts

Run the deploy.yml playbook against the webservers group Do a dry run of the hardening playbook to see what would change List all hosts in the production inventory Ping all hosts in the staging group and show results

Alerting โ€” PagerDuty & OpsGenie

DevOps Agent supports 16 alerting tools โ€” 8 for PagerDuty and 8 for OpsGenie โ€” for incident management, on-call schedules, and alert routing.

PagerDuty (8 tools)

Set PAGERDUTY_API_KEY in your environment. Requires a REST API key from PagerDuty's API Access section.

CategoryOperations
IncidentsList, get details, create, acknowledge, resolve
On-callGet current on-call users for a schedule
ServicesList PagerDuty services
EscalationList escalation policies

OpsGenie (8 tools)

Set OPSGENIE_API_KEY in your environment.

CategoryOperations
AlertsList, get details, create, acknowledge, close
On-callGet current on-call users
Teams & SchedulesList teams, list schedules

Example prompts

Show all open PagerDuty incidents with high urgency Acknowledge incident P-12345 in PagerDuty Who is currently on call for the payments schedule? Create an OpsGenie alert: Database connection pool exhausted on prod-db-01 Close all acknowledged OpsGenie alerts from the last hour

Tools Overview

The agent has access to 727 tools across 17 categories. You never call tools directly โ€” the agent selects and chains the right ones based on your request.

Provider / CategoryToolsServices covered
AWS334EC2, VPC, IAM, RDS, Lambda, ECS, EKS, S3, DynamoDB, CloudFormation, Route53, SNS/SQS, CloudWatch, ElastiCache, EFS, EventBridge, Kinesis, NAT, ALB, ECR, Secrets Manager
Azure38VMs, Storage, SQL, NSGs, Container Instances, Disks, VNets, Public IPs
GCP42Compute, GCS, Cloud SQL, GKE, Cloud Run, Pub/Sub, Secret Manager, BigQuery, Firewall, Cloud Functions, Logs
Kubernetes76Pods, Deployments, Services, Namespaces, ConfigMaps, Secrets, Ingress, Nodes, Jobs
CI/CD35Jenkins (jobs, builds, pipelines, queue), GitHub Actions (workflows, runs, secrets)
Docker22Containers (lifecycle, exec, logs), Images, Volumes, Networks, Compose
Terraform25Init, plan, apply, destroy, state management, workspaces, outputs
SonarQube12Projects, issues, quality gates, coverage & technical debt metrics
Nexus12Repositories, components, assets, system health
Ansible13Playbooks, inventory, ad-hoc commands, roles, Galaxy
Monitoring12Prometheus, CloudWatch metrics, threshold alerts
Alerting16PagerDuty (incidents, on-call, services), OpsGenie (alerts, teams, schedules)
Pentest12Port scanning, SSL checks, HTTP security headers, DNS enumeration
Git6Status, log, diff, branches, commits, stash

Safety levels

Every tool is classified into one of three risk levels:

LevelExamplesBehaviour
LowList, describe, get, read operationsExecutes immediately
MediumUpdate, restart, scale, IAM changesAgent asks for confirmation
HighDelete, terminate, drop, purge operationsRequires explicit yes/no confirmation. Escalates to Critical if targeting production.

Monitoring & Alerts

The Monitoring tab lets you create threshold-based alerts on any metric exposed by your cloud provider or Prometheus.

Creating a monitor

1
Go to Monitoring โ†’ New Rule
Choose a metric (CPU, memory, error rate, latency, disk, etc.).
2
Set the threshold
Enter a numeric value (e.g. 85 for 85% CPU). The threshold must be a number โ€” text values are rejected.
3
Choose notification channels
Email, Slack webhook, or PagerDuty. Alerts fire when the rule enters firing state and again when it resolves.

Incident Response

The Incidents tab provides automated runbook execution. When an alert fires, you can trigger a playbook that diagnoses and remediates the issue automatically.

Built-in playbooks

PlaybookWhat it does
high-cpu-usageIdentifies the top CPU-consuming processes, optionally restarts the service or scales the deployment
memory-pressureChecks for memory leaks, triggers a restart or node cordon if critical
disk-space-lowLists large files and stale logs, creates a cleanup plan
pod-crashloopFetches pod logs, identifies the crash reason, applies a fix or rolls back
service-downRuns connectivity checks, restarts the service, escalates if unresolved
security-alertBlocks the offending IP via security group rule, notifies the team
โ„น๏ธ You can also trigger any playbook manually from the Incidents tab โ†’ Run Playbook, without waiting for an alert to fire.
Developer

REST API

The DevOps Agent REST API lets you integrate cloud automation directly into your own tools, scripts, and CI/CD pipelines. All endpoints require an API key โ€” generate one from Settings โ†’ Developer โ†’ API Keys.

Authentication

Pass your API key in the Authorization header on every request:

Authorization: Bearer YOUR_API_KEY

Alternatively, use the api_key query parameter for clients that cannot set headers:

GET /api-gateway/v1/resources?api_key=YOUR_API_KEY

Base URL

https://devopsagent.io/api-gateway/v1

Endpoints

MethodPathDescription
GET/resourcesList all synced infrastructure resources
GET/resources/:idGet a specific resource by ID
GET/analytics/costsCost analytics for the current billing period
GET/incidentsList all incidents
POST/incidentsCreate a new incident manually
GET/healthHealth check โ€” no auth required, for load balancers

Rate limiting

Requests are rate-limited per API key. The current limit and remaining quota are returned on every response:

HeaderMeaning
X-RateLimit-LimitMaximum requests allowed per hour
X-RateLimit-RemainingRequests remaining in the current window
X-RateLimit-ResetUnix timestamp when the window resets

When the limit is exceeded the API returns 429 Too Many Requests.

Example โ€” list resources

curl https://devopsagent.io/api-gateway/v1/resources \
  -H "Authorization: Bearer YOUR_API_KEY"
{
  "resources": [
    { "id": 1, "name": "prod-web-01", "type": "ec2", "status": "running", "region": "us-east-1" },
    { "id": 2, "name": "prod-db",     "type": "rds", "status": "available", "region": "us-east-1" }
  ]
}

MCP Server

DevOps Agent exposes all cloud tools as an MCP (Model Context Protocol) server registered on the official MCP Registry. Connect any MCP-compatible AI client โ€” Claude Desktop, Claude.ai, or custom agents โ€” and manage your infrastructure without leaving your AI assistant.

Registry

The server is published on the official MCP Registry and npm:

RegistryIdentifier
MCP Registryio.github.shittuay/devops-agent
npm@devopsagent/mcp-server

Connect via Claude.ai (Connectors)

Go to Claude.ai โ†’ Settings โ†’ Connectors โ†’ Add custom connector and fill in:

FieldValue
NameDevOps Agent
Remote MCP server URLhttps://devopsagent.io/mcp
OAuth Client IDclaude-connector
OAuth Client Secretdevopsagent-mcp-secret-2026

Click Add. Claude will open a DevOps Agent login page โ€” sign in with your DevOps Agent credentials to authorize the connection. All your tools will be available in Claude immediately after.

Connect via Claude Desktop

Generate an API key at /api-gateway/keys, then edit your claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%\Claude\):

{
  "mcpServers": {
    "devops-agent": {
      "url": "https://devopsagent.io/mcp/sse",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Restart Claude Desktop โ€” all DevOps Agent tools appear in the tool list automatically.

Endpoints

MethodPathDescription
POST/mcpStreamable HTTP transport (MCP 2025-03-26). Used by Claude.ai connectors. Returns JSON-RPC responses inline.
GET/mcp/sseHTTP+SSE transport. Used by Claude Desktop. Server sends an endpoint event with your session's message URL.
POST/mcp/messages?sessionId=โ€ฆJSON-RPC message endpoint for SSE sessions. Responses are pushed back via the SSE stream.
GET/mcp/healthPublic health check. Returns tool count and protocol version. No auth required.

Authentication

Two authentication methods are supported:

MethodHowUsed by
OAuth 2.0Authorization Code flow with PKCE. Client ID + Secret in connector settings.Claude.ai connectors
API KeyAuthorization: Bearer YOUR_API_KEY header, or ?api_key= query param.Claude Desktop, custom clients

Generate API keys at /api-gateway/keys.

OAuth 2.0 Endpoints

PathDescription
/.well-known/oauth-protected-resourceResource metadata (RFC 9728) โ€” advertises the authorization server URL.
/.well-known/oauth-authorization-serverAuthorization server metadata (RFC 8414) โ€” lists all OAuth endpoints.
/mcp/oauth/authorizeAuthorization endpoint โ€” shows login page, issues auth code on success.
/mcp/oauth/tokenToken endpoint โ€” exchanges auth code for access token. Supports client_secret_post and client_secret_basic.

Supported JSON-RPC methods

MethodDescription
initializeHandshake โ€” returns server name, version, and capabilities
tools/listReturns all tools with names, descriptions, and input schemas
tools/callExecute a tool by name with the provided arguments
pingKeepalive โ€” server returns an empty result

Example โ€” streamable HTTP (Claude.ai)

POST https://devopsagent.io/mcp
Authorization: Bearer YOUR_ACCESS_TOKEN
Content-Type: application/json

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_ec2_instances",
    "arguments": { "region": "us-east-1" }
  }
}
{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "content": [{ "type": "text", "text": "{ \"success\": true, \"instances\": [...] }" }],
    "isError": false
  }
}
โ„น๏ธ All tool calls go through the same safety validator as the chat interface. High-risk operations (delete, terminate) include a warning in the response so the calling model can re-confirm with the user.

Protocol

Implements MCP protocol version 2025-03-26. SSE sessions expire after 10 minutes of inactivity with keepalive comments every 15 seconds. OAuth access tokens are valid for 30 days.

Billing & Plans

DevOps Agent is billed per subscription tier. Each tier includes a monthly credit allocation that refreshes on your billing date.

PlanPriceCredits/moAI Model
Free$020Kimi K2.5
Starter$29/mo1,000Kimi K2.5
Professional$99/mo5,000Claude Sonnet
Business$299/mo10,000Claude Opus

Credit packs

Need more credits without upgrading your plan? Purchase one-time credit packs from the Billing page:

PackPrice (USD)Price (NGN)
100 credits$10โ‚ฆ5,000
250 credits$20โ‚ฆ9,000
500 credits$35โ‚ฆ15,000
1,000 credits$60โ‚ฆ25,000

Payments are processed securely via Stripe. Both USD and NGN currencies are supported โ€” the currency is auto-detected from your location and can be manually toggled on the billing page.

Security & 2FA

Two-factor authentication

Enable 2FA in Settings โ†’ Security โ†’ Two-Factor Authentication. DevOps Agent supports both:

MethodHow to set up
Authenticator App (TOTP)Scan the QR code with Google Authenticator, Authy, or 1Password. Save your backup codes in a safe place.
Email OTPA one-time code is sent to your registered email address on each login.

Lost access to your authenticator?

If you can no longer access your authenticator app (deleted the app, got a new phone, etc.), you can reset 2FA from the login screen:

1
Enter your email and password
Log in as normal. When the 2FA verification screen appears, look for the "Lost access to your authenticator?" link at the bottom.
2
Click the link
A reset link is sent to your registered email address. It expires in 30 minutes.
3
Click the link in your email
2FA is disabled and you are logged in automatically. You can then re-configure 2FA in Settings โ†’ Security.
Security note: The reset link is single-use and tied to your account. If you did not request it, contact support immediately.

Credential encryption

All cloud credentials (AWS keys, Azure client secrets, GCP service account JSON) are encrypted at rest using AES-128 (Fernet) before being stored in the database. Credentials are decrypted only at request time, in memory, and never logged.

Guardrails

Every message passes through a three-layer guardrail stack before any credits are charged:

1
Pattern matching
10 regex rules block prompt injection, jailbreak attempts, and out-of-scope requests instantly, before any AI call is made.
2
AI pre-screen
A lightweight AI model classifies the intent of the message and blocks social engineering, malware requests, and data exfiltration attempts.
3
Hardened system prompt
The main agent has a fixed identity and cannot be convinced to reveal its instructions, generate malware, or impersonate operators.

Teams

The Teams tab is available on Starter plans and above. It lets an admin invite colleagues to the same workspace so they share access to cloud integrations and conversations.

Roles

RoleCan do
AdminInvite/remove members, change roles, manage cloud credentials, full agent access
ApproverReview and approve high-risk operations flagged by the safety system
UserFull agent access, cannot manage team or credentials
ViewerRead-only access to conversations and dashboard
โ„น๏ธ Each user has fully isolated conversations, cloud credentials, and credit balances โ€” even within the same team workspace. One user cannot see another user's chat history.

Settings

SettingWhereWhat it does
Cloud CredentialsSettings โ†’ CloudAdd or update AWS, Azure, GCP credentials. Each provider has a Test Connection button.
Learning ModeSettings โ†’ PreferencesAgent explains every action inline. Tracks proficiency per topic.
Sandbox ModeSettings โ†’ PreferencesAll tool calls return simulated results. No real cloud calls made.
Experience LevelSettings โ†’ PreferencesBeginner / Intermediate / Advanced โ€” controls explanation depth in Learning Mode.
Model OverrideChat toolbarForce Kimi or Claude regardless of your subscription tier default.
Two-Factor AuthSettings โ†’ SecurityEnable TOTP or email OTP for login.
PasswordSettings โ†’ SecurityChange your login password.
NotificationsSettings โ†’ NotificationsConfigure email alerts for monitoring rules and billing events.

FAQ

Can the agent accidentally delete my cloud resources?

All delete and terminate operations require an explicit confirmation from you before executing. The agent will describe exactly what will be deleted and wait for your yes/no. You can always type "no" or "cancel" to abort.

Where are my cloud credentials stored?

Credentials are stored encrypted in our database using AES-128 encryption. They are decrypted in memory only when a tool call needs them, and are never written to logs or included in AI model context.

What happens when I run out of credits?

On the Free plan, you'll see an upgrade prompt when credits reach zero โ€” no charge is made automatically. On paid plans, your subscription auto-renews monthly. You can also purchase credit packs any time from the Billing page.

Does the agent have access to my data inside cloud resources (e.g. database contents, S3 file contents)?

No. The agent can list, describe, start, stop, and manage the lifecycle of cloud resources โ€” but it does not read the contents of databases, S3 objects, or secrets stored in Secrets Manager unless you explicitly paste content into the chat.

What regions are supported for AWS?

All AWS regions. The default region is set when you configure your credentials. You can override the region per-message by including it in your prompt, e.g. list EC2 instances in eu-west-1.

How do I cancel my subscription?

Go to Billing โ†’ Manage Subscription โ†’ Cancel plan. Your subscription stays active until the end of the current billing period, then downgrades to the Free plan. No refunds are issued for unused credits in the current period.

Is there an API I can call programmatically?

Not yet. API access for programmatic integration is on the roadmap. Watch the GitHub repo for updates.