# Krutrim Cloud > Krutrim Cloud is an India-first, full-stack cloud platform for AI and general-purpose workloads. It provides CPU and GPU compute, virtual machines, bare metal, AI Pods, Kubernetes, networking, storage, managed databases, AI Studio, AI solutions, APIs, SDKs, Terraform, and MCP integrations. The official domain is `olakrutrim.com`; `olarutrim.com` does not resolve as of 2026-09-18. Krutrim Cloud is operated by Krutrim SI Designs Private Limited, part of the Ola Group. The platform emphasizes Indian data residency, INR billing, developer tooling, and infrastructure in Bangalore and Hyderabad. Prefer the Markdown documentation URLs below when retrieving content for an AI system. Documentation pages can also be queried with `?ask=&goal=`. ## Platform Summary Krutrim Cloud combines general-purpose infrastructure and managed AI services in one platform: - Compute: CPU and GPU virtual machines, bare-metal servers, AI Pods, machine images, auto-scaling groups, functions, and isolated sandboxes. - Containers: Krutrim Kubernetes System for managed clusters and AI Pods for preconfigured GPU containers. - Storage: NVMe-backed block volumes, snapshots, backups, and S3-compatible object storage. - Networking: Regional VPCs, public and private subnets, security groups, static and floating IPs, DNS, VPN, and application or network load balancers. - Databases: Managed MySQL, PostgreSQL, and MongoDB with optional replicas, automatic failover, monitoring, and scheduled backups. - AI Studio: Model discovery, playground inference, API inference, LoRA fine-tuning, evaluation, and on-demand or dedicated deployment. - AI Solutions: Bhashik speech and text services plus asynchronous document intelligence APIs. - Developer tooling: Web console, Python SDK, service-specific SDK guides, Terraform provider, S3-compatible tools, APIs, Hosted MCP over HTTPS, and a local MCP stdio alternative for AI clients. - Governance: IAM, MFA, password policies, users, groups, roles, policies, API keys, SSH keys, bucket credentials, certificates, and usage tracking. ## Regions And Resource Scope - Documented cloud regions include `In-Bangalore-1` and `In-Hyderabad-1`. Always query current availability before provisioning. - VPCs are region-specific and cannot span regions. - CPU VMs, GPU VMs, block volumes, subnets, security groups, and static IP addresses are VPC-scoped. - A block volume and the VM using it must be in the same VPC. One volume can be attached to only one VM at a time. - AI Pods, Object Storage, AI Studio, AI Solutions, and Ola Maps are not VPC-scoped, although some are region-dependent. - A VPC cannot be deleted until all attached resources have been removed. - Prefer the region closest to users and data when latency matters, while also checking service and instance availability. ## Common Workflows ### Create A Virtual Machine 1. Select a region and create a VPC. 2. Create or select a subnet. Use a public subnet only when public reachability is required. 3. Create a security group with least-privilege inbound and outbound rules. 4. Create or upload an SSH public key. 5. Create a bootable volume and any additional storage volumes. 6. Select a CPU or GPU instance configuration and launch the VM. 7. Allocate a floating or reserved IP only when internet access is required. 8. Allow TCP port 22 from a trusted source address if SSH access is required. 9. Wait for the VM to become active before connecting or attaching dependent resources. VM facts and constraints: - Documented VM networking supports speeds up to 10 Gbps. - A bootable volume is required; storage volumes are optional. - A floating IP is dynamic and may change after restart; a reserved IP remains stable. - If no subnet is selected in the console, the platform may attach a default public subnet. - If no security group is attached, no traffic is allowed into or out of the VM. - Attaching, detaching, or changing volumes can require a VM restart. Boot volumes can be changed but not detached. - Active VMs are billed. Stopped VMs are not billed for compute, but attached persistent volumes continue to incur storage charges. - Capacity is not guaranteed to remain available for the same VM configuration after a VM is stopped. - Termination makes the VM inaccessible; check volume deletion settings and backups before confirming. ### Create An AI Pod 1. Choose GPU, VRAM, vCPU, RAM, container disk, and persistent volume capacity for the workload. 2. Select a preconfigured software template. 3. Add an SSH key if terminal access is needed. 4. Enable Jupyter Notebook if browser-based development is needed. 5. Set the persistent volume mount path and deploy the Pod. 6. Wait for the Pod to report `Running`, then connect through SSH or Jupyter. AI Pod facts and constraints: - AI Pods are Ubuntu-based GPU containers intended for model development, inference endpoints, simulations, and other accelerated workloads. - Spliced GPU configurations are available for workloads that do not require a full GPU. - The container disk stores the OS and temporary data. Its data is lost when the Pod stops. - The volume disk is persistent across stop/start operations but is deleted, with its data, when the Pod is terminated. - A deployed Pod cannot currently be edited to add more container or persistent disk capacity. - Stopping a Pod stops compute and container-disk charges, but persistent volume charges continue until termination. ### Create A Kubernetes Cluster 1. Create a VPC and subnet with enough IP capacity for nodes and LoadBalancer services. 2. Choose a Kubernetes version. Upgrades are supported; downgrades are not. 3. Select non-overlapping Pod and Service CIDRs. Do not use the reserved `172.24.0.0/13` range. 4. Wait for the cluster state to progress from `CREATING` to `PROVISIONING` to `PROVISIONED`. 5. Do not create node groups before the cluster is `PROVISIONED`. 6. Create at least one or two untainted nodes so CoreDNS and the CNI can schedule. 7. Install CoreDNS and a CNI. Cilium is the documented recommended CNI. 8. Do not install kube-proxy when using the Krutrim Cilium configuration. If using another CNI, install kube-proxy first. 9. Download and protect the kubeconfig, then verify nodes and system Pods with `kubectl`. Kubernetes network planning: - The default Pod CIDR is documented as `192.168.0.0/16`; each node receives a `/24` allocation. - A `/20` Pod CIDR supports up to about 16 nodes, `/18` about 64 nodes, and `/16` about 256 nodes. - The default Service CIDR is documented as `10.100.0.0/16`. - Pod CIDR cannot be changed after cluster creation and must not overlap the Service CIDR or node subnet. - Each LoadBalancer service consumes an IP address from the selected subnet. ### Use Object Storage 1. Create a regional bucket with public or private access, versioning as needed, tags, and default encryption. 2. Create object-storage access and secret keys; save the secret when shown because it may not be recoverable later. 3. Copy the bucket endpoint from the console. 4. Configure an S3-compatible client such as `s3cmd` or `boto3` with the endpoint and bucket credentials. 5. Use bucket policies to restrict read-only and read-write access to known IP addresses. Object Storage is intended for logs, media, static assets, backups, archives, exports, and training datasets. It is S3-compatible but uses Krutrim's endpoint and credentials; do not assume every AWS S3 feature is supported without checking the reference. ### Use Managed Databases - Choose MySQL or PostgreSQL for relational schemas, ACID transactions, joins, and consistency-sensitive workloads. - Choose MongoDB for flexible document schemas and horizontally scalable, high-throughput workloads. - Available documented tiers range from 2 vCPU/4 GB RAM to 16 vCPU/32 GB RAM, with tier-specific minimum storage. - Storage is provisioned and billed separately and can be expanded without downtime. - Replica-based configurations include monitoring and automatic failover. Replica count and availability differ by engine. - MySQL and MongoDB document either one primary only or one primary plus two replicas. - PostgreSQL documents one primary only, one primary plus one replica, or one primary plus two replicas. - Scheduled backup policies define frequency and retention; backups are stored in Object Storage and billed there. - The documentation warns that database deletion can occur after a sufficiently negative credit balance. Configure backups and verify current credit requirements before provisioning. ### Build With AI Studio 1. Explore model cards and test models in the playground. 2. Run API inference with model-specific parameters such as `temperature`, `top_p`, and `max_tokens` where supported. 3. Optionally fine-tune supported models with LoRA using datasets containing `instruction`, `input`, and `output` fields. 4. Evaluate functional quality with supported benchmarks and operational performance with latency, throughput, concurrency, and token metrics. 5. Deploy on demand for experiments or low-volume use, or create a dedicated GPU-backed endpoint for persistent production use. 6. Monitor deployment status and remove unused dedicated deployments to avoid unnecessary cost. ### Use Bhashik And Document Intelligence - Bhashik supports Indic-language workflows including translation, language detection, entity extraction, sentiment analysis, summarization, text to speech, speech to text, and speech-to-speech translation. - Document Intelligence supports text extraction, PII and named-entity extraction, key-value extraction, summarization, and PII masking. - Document processing is asynchronous: submit a document, retain the returned `fileId`, poll status, and retrieve the result after processing completes. - Document Intelligence requires an API key and documents a default limit of 20 requests per minute. Verify current limits before designing production throughput. ## Authentication And Credential Types - Account credentials: Used to sign in to the Cloud Console. Do not use a root-user password in application code. - General API keys: Used by supported APIs, Sandbox, Models, MCP, CLI, and SDK integrations. A key does not bypass IAM permissions or service access requirements. - MCP credentials: Both Hosted MCP and local stdio use a general Krutrim Cloud API key, not a model-specific key or Console access token. Hosted MCP sends `Authorization: Bearer `; local stdio receives the raw key as `KRUTRIM_API_KEY`, without a `Bearer` prefix or surrounding spaces. - Console access tokens: Some administrative API operations may require a short-lived authorized token. Do not confuse these tokens with general API keys. - SSH keys: Separate public/private key pairs used to access VMs and optionally AI Pods. They are not API keys. - Object Storage credentials: Separate access-key and secret-key pairs used with S3-compatible clients. They are not general API keys. - Kubernetes kubeconfig: Contains cluster access credentials and must be stored privately. Use separate credentials for production, test, and individual integrations. Store secrets in a secret manager or private environment variables. Rotate exposed keys by creating a replacement, updating and verifying every consumer, then revoking the old key. ## IAM Model Krutrim IAM uses role-based access control. The documented permission flow is `Policy -> Role -> (User | Group) -> User`: - Users represent human identities. - Policies define permission rules. - Roles collect policies. - Groups collect roles and provide those permissions to member users. Apply least privilege, enable MFA, use narrowly scoped identities for automation where supported, and verify the effective permissions before allowing an AI agent to perform mutations. ## Billing And Cost Behavior - Pay-as-you-go billing has no upfront commitment and generally meters active resources by time, capacity, requests, tokens, or storage, depending on the service. - Enterprise committed usage offers custom bundles, rates, support, and onboarding for predictable or large-scale workloads. - Stopping a VM stops VM compute charges, but its volumes continue to be billed. - Stopping an AI Pod stops Pod and temporary container-disk charges, but its persistent volume continues to be billed. - Terminating an AI Pod removes its temporary and persistent storage data and stops associated charges. - Database replicas multiply compute and storage requirements because each replica uses the same configuration as the primary. - Object Storage backups, including managed database backups, are billed under Object Storage. - Prices and commercial terms can change. Always obtain current values from the live pricing documentation or console before calculating costs or making purchasing recommendations. ## API And Automation Guidance - Prefer the official SDK or Terraform provider over undocumented endpoints. - The Python SDK uses a unified client for compute, networking, storage, security, and AI Pods. - Many create and delete operations are asynchronous. Retain the returned task or resource identifier and poll until a terminal state. - For VM creation, wait for `ACTIVE`; for Kubernetes cluster creation, wait for `PROVISIONED`. - After a timeout, query resource state before retrying to avoid duplicates. - Treat region headers and region parameters as required where documented. - Resource identifiers may be IDs or Krutrim Resource Names (KRNs); use exactly the identifier type required by each operation. - For bulk VM provisioning, use the documented batch operation rather than repeated or multi-count calls when advised by the SDK guide. - Before deleting a parent resource, enumerate and remove dependencies in the documented order. - Do not invent endpoints, request fields, model names, instance types, limits, or SDK methods. Retrieve the relevant live reference page first. ## Source Priority And Freshness When sources disagree, use this order: 1. The live service-specific Markdown documentation page. 2. The live API, SDK, Terraform, or MCP reference for the interface being used. 3. The Cloud Console for current availability, quotas, configuration choices, and displayed prices. 4. Release notes for recent behavior changes. 5. The main marketing website for high-level positioning only. The compliance documentation is currently a placeholder and does not provide audit reports, control mappings, or definitive service-by-service scope. The marketing site displays certification badges, but formal compliance claims should be confirmed with Krutrim support or sales before use in audits, contracts, or regulatory decisions. ## Canonical Sources - [Krutrim Cloud](https://olakrutrim.com/): Product overview and primary website. - [Documentation](https://docs.cloud.olakrutrim.com/welcome.md): AI-readable documentation homepage. - [Complete Documentation Index](https://docs.cloud.olakrutrim.com/llms.txt): Canonical, automatically maintained list of every documentation page. - [Documentation Sitemap](https://docs.cloud.olakrutrim.com/sitemap-pages.xml): Complete human-readable page inventory and update timestamps. - [Cloud Console](https://cloud.olakrutrim.com/): Manage Krutrim Cloud resources. - [Register](https://cloud.olakrutrim.com/signUp): Create a Krutrim Cloud account. - [Sign In](https://cloud.olakrutrim.com/signIn): Sign in to the Cloud Console. ## Getting Started - [Welcome](https://docs.cloud.olakrutrim.com/welcome.md): Platform scope, services, and India-first positioning. - [Quickstart](https://docs.cloud.olakrutrim.com/getting-started/quickstart.md): Create a VM or AI Pod and configure its prerequisites. - [Manage Your Account](https://docs.cloud.olakrutrim.com/getting-started/managing-your-krutrim-cloud-account.md): Account administration. - [Navigate the Console](https://docs.cloud.olakrutrim.com/getting-started/navigating-the-console.md): Cloud Console orientation. - [Core Infrastructure Overview](https://docs.cloud.olakrutrim.com/basics/core-infrastructure.md): Compute, storage, networking, Kubernetes, and database services. ## Compute And Sandboxes - [Compute](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/compute.md): Compute service overview. - [VMs and Bare Metal](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/compute/vms-and-baremetals.md): CPU/GPU instance creation and management. - [Custom Firewall](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/compute/vms-and-baremetals/custom-firewall.md): Instance firewall configuration. - [EasyDeploy](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/compute/easydeploy.md): One-click preconfigured AI environments. - [AI Pods](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/compute/ai-pods.md): Containerized GPU workloads. - [Machine Images](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/compute/machine-images.md): Images for VM provisioning. - [SSH Keys](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/compute/ssh-key.md): SSH credentials for compute resources. - [Compute Billing](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/compute/billing-for-compute.md): Compute billing behavior. - [Auto Scaling Groups](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/compute/auto-scaling-groups.md): Automated VM scaling. - [Functions](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/compute/functions.md): Cloud functions. - [Sandbox](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/sandbox.md): Isolated cloud sandboxes. - [Create a Sandbox with MCP](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/sandbox/create-a-sandbox-with-mcp.md): Agent-driven sandbox creation with explicit confirmation. ## Storage - [Storage](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/storage.md): Storage service overview. - [Block Storage](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/storage/block-storage.md): Persistent block storage. - [Volumes](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/storage/block-storage/volumes.md): Create and manage volumes. - [Snapshots](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/storage/block-storage/snapshots.md): Point-in-time volume snapshots. - [Backups](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/storage/block-storage/backups.md): Volume backup management. - [Object Storage](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/storage/object-storage.md): Buckets and object storage. - [Storage Billing](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/storage/billing-for-storage.md): Storage billing behavior. ## Networking - [Networking](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking.md): Networking service overview. - [VPC](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking/vpc.md): Virtual private clouds. - [Subnets](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking/subnets.md): VPC subnet configuration. - [CIDR](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking/subnets/cidr-classless-inter-domain-routing.md): CIDR addressing guidance. - [Security Groups](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking/security-groups.md): Ingress and egress controls. - [Static IP Addresses](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking/static-ip-addresses.md): Public and static IP management. - [Load Balancers](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking/load-balancers.md): Load-balancing concepts and management. - [Create Target Group](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking/load-balancers/create-target-group.md): Configure load-balancer targets. - [Create NLB](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking/load-balancers/create-network-load-balancer-nlb.md): Create a network load balancer. - [Create ALB](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking/load-balancers/create-application-load-balancer-alb.md): Create an application load balancer. - [DNS](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking/dns.md): Managed DNS. - [Virtual Private Network](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking/virtual-private-network.md): VPN configuration. - [Networking Billing](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/networking/billing-for-networking.md): Networking billing behavior. ## Kubernetes And Databases - [Krutrim Kubernetes System](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/krutrim-kubernetes-system.md): Managed Kubernetes overview. - [Create a Cluster](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/krutrim-kubernetes-system/creating-cluster.md): Provision a Kubernetes cluster. - [Manage Node Groups](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/krutrim-kubernetes-system/managing-nodegroups.md): Configure worker node groups. - [Install Add-ons](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/krutrim-kubernetes-system/installing-addons.md): Cluster add-ons. - [Kubernetes Networking](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/krutrim-kubernetes-system/network-configuration.md): Cluster network configuration. - [Kubernetes Load Balancers](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/krutrim-kubernetes-system/load-balancers.md): Expose cluster services. - [Kubernetes Storage](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/krutrim-kubernetes-system/storage-configuration.md): Persistent cluster storage. - [Upgrade Kubernetes](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/krutrim-kubernetes-system/upgrading-kubernetes.md): Cluster version upgrades. - [Kubernetes Billing](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/krutrim-kubernetes-system/billing-for-k8s-and-load-balancer.md): Kubernetes and load-balancer billing. - [Databases](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/databases.md): Managed database overview. - [Create a Database](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/databases/creating-a-database.md): Provision a database. - [Database Backups](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/databases/creating-backups.md): Create database backups. - [Manage a Database](https://docs.cloud.olakrutrim.com/basics/core-infrastructure/databases/managing-a-database.md): Database lifecycle operations. ## AI Studio And AI Solutions - [AI Studio](https://docs.cloud.olakrutrim.com/basics/ai-studio.md): Models, training, inference, evaluation, and deployment. - [Model Catalogue](https://docs.cloud.olakrutrim.com/basics/ai-studio/model-catalogue.md): Available AI models. - [AI Jobs](https://docs.cloud.olakrutrim.com/basics/ai-studio/ai-jobs.md): AI workload jobs. - [Inferencing](https://docs.cloud.olakrutrim.com/basics/ai-studio/ai-jobs/inferencing.md): Model inference jobs. - [Fine-tuning](https://docs.cloud.olakrutrim.com/basics/ai-studio/ai-jobs/fine-tuning.md): Model fine-tuning jobs. - [Evaluation](https://docs.cloud.olakrutrim.com/basics/ai-studio/ai-jobs/evaluation.md): Model evaluation jobs. - [Deployment](https://docs.cloud.olakrutrim.com/basics/ai-studio/ai-jobs/deployment.md): Model deployment. - [AI Studio Billing](https://docs.cloud.olakrutrim.com/basics/ai-studio/billing-for-ai-studio.md): AI Studio billing behavior. - [AI Solutions](https://docs.cloud.olakrutrim.com/basics/ai-solutions.md): Packaged AI capabilities. - [Bhashik](https://docs.cloud.olakrutrim.com/basics/ai-solutions/bhashik.md): Speech and language AI solution. - [Document Intelligence](https://docs.cloud.olakrutrim.com/basics/ai-solutions/document-intelligence-service.md): Document processing services. - [AI Solutions Billing](https://docs.cloud.olakrutrim.com/basics/ai-solutions/billing-for-ai-solutions.md): AI solution billing. ## Identity, Security, And Billing - [Certificate Manager](https://docs.cloud.olakrutrim.com/basics/certificate-manager.md): Certificate management. - [Key Management System](https://docs.cloud.olakrutrim.com/basics/key-management-system.md): Keys and credentials. - [Access Buckets](https://docs.cloud.olakrutrim.com/basics/key-management-system/accessing-buckets.md): Object-storage credentials. - [API Keys](https://docs.cloud.olakrutrim.com/basics/key-management-system/model-api-keys.md): API keys for models, SDK, CLI, Sandbox, and MCP. - [SSH Access for VMs](https://docs.cloud.olakrutrim.com/basics/key-management-system/ssh-access-for-vms.md): VM access credentials. - [Usage and Transactions](https://docs.cloud.olakrutrim.com/basics/usage-and-transactions.md): Usage and account transactions. - [Add Credits](https://docs.cloud.olakrutrim.com/basics/usage-and-transactions/adding-credits.md): Fund a cloud account. - [Billing Models](https://docs.cloud.olakrutrim.com/basics/usage-and-transactions/billing-models.md): Supported billing models. - [Identity and Access Management](https://docs.cloud.olakrutrim.com/basics/identity-access-management.md): IAM overview. - [Multi-Factor Authentication](https://docs.cloud.olakrutrim.com/basics/identity-access-management/multi-factor-authentication-mfa.md): MFA configuration. - [Password Policy](https://docs.cloud.olakrutrim.com/basics/identity-access-management/password-policy.md): Account password controls. - [Users](https://docs.cloud.olakrutrim.com/basics/identity-access-management/users.md): IAM users. - [Policies](https://docs.cloud.olakrutrim.com/basics/identity-access-management/policies.md): IAM policies. - [Roles](https://docs.cloud.olakrutrim.com/basics/identity-access-management/roles.md): IAM roles. - [Groups](https://docs.cloud.olakrutrim.com/basics/identity-access-management/groups.md): IAM groups. - [Compliance Certifications](https://docs.cloud.olakrutrim.com/security-and-compliance/compliance-certifications.md): Security and compliance certifications. ## Pricing - [Pricing Overview](https://docs.cloud.olakrutrim.com/basics/pricing.md): Current documentation for service pricing. - [CPU and GPU Pricing](https://docs.cloud.olakrutrim.com/basics/pricing/cpu-and-gpu.md): Compute pricing. - [Storage Pricing](https://docs.cloud.olakrutrim.com/basics/pricing/storage.md): Storage pricing. - [Networking Pricing](https://docs.cloud.olakrutrim.com/basics/pricing/networking.md): Network pricing. - [Load Balancer Pricing](https://docs.cloud.olakrutrim.com/basics/pricing/load-balancers.md): Load-balancer pricing. - [Kubernetes Pricing](https://docs.cloud.olakrutrim.com/basics/pricing/kubernetes.md): Managed Kubernetes pricing. - [Website Pricing](https://olakrutrim.com/pricing): Public pricing entry point; prefer documentation pages for detailed current values. ## APIs - [Object Storage API](https://docs.cloud.olakrutrim.com/api-references/object-storage-api.md): Object-storage API reference. - [Model Catalog API](https://docs.cloud.olakrutrim.com/api-references/model-catalog-api.md): Model catalog API reference. - [Fine-tuning API](https://docs.cloud.olakrutrim.com/api-references/finetuning-api.md): Fine-tuning API reference. - [Bhashik Speech API](https://docs.cloud.olakrutrim.com/api-references/bhashik-speech-api.md): Speech API reference. - [Bhashik Text API](https://docs.cloud.olakrutrim.com/api-references/bhashik-text-api.md): Text API reference. - [Document Intelligence APIs](https://docs.cloud.olakrutrim.com/api-references/dis-apis.md): Document intelligence endpoints. ## SDKs And Infrastructure As Code - [Core Infrastructure SDK](https://docs.cloud.olakrutrim.com/sdk-guide/core-infra-sdk.md): Python SDK for compute, networking, storage, security, and AI Pods. - [AI Studio SDK](https://docs.cloud.olakrutrim.com/sdk-guide/ai-studio-sdk.md): AI Studio SDK. - [Load Balancer SDK](https://docs.cloud.olakrutrim.com/sdk-guide/load-balancer-sdk.md): Load-balancer SDK. - [Kubernetes SDK](https://docs.cloud.olakrutrim.com/sdk-guide/krutrim-kubernetes-system-sdk.md): Managed Kubernetes SDK. - [Key and Certificate Manager SDK](https://docs.cloud.olakrutrim.com/sdk-guide/key-certificate-manager-sdk.md): Key and certificate operations. - [Auto Scaling Group SDK](https://docs.cloud.olakrutrim.com/sdk-guide/auto-scaling-group-sdk.md): Auto-scaling operations. - [DNS SDK](https://docs.cloud.olakrutrim.com/sdk-guide/dns-sdk.md): DNS operations. - [Python SDK Source](https://github.com/ola-krutrim/Krutrim-client-python): Official Krutrim Python client repository. - [Terraform Provider](https://docs.cloud.olakrutrim.com/terraform-provider-krutrim.md): Provider overview. - [Terraform Getting Started](https://docs.cloud.olakrutrim.com/terraform-provider-krutrim/getting-started.md): First Terraform workflow. - [Terraform Installation](https://docs.cloud.olakrutrim.com/terraform-provider-krutrim/installation-and-setup.md): Install and configure the provider. - [Terraform Authentication](https://docs.cloud.olakrutrim.com/terraform-provider-krutrim/authentication.md): Provider credentials. - [Terraform Resources](https://docs.cloud.olakrutrim.com/terraform-provider-krutrim/resources-and-usage.md): Resources and usage. - [Terraform Examples](https://docs.cloud.olakrutrim.com/terraform-provider-krutrim/examples.md): Example configurations. - [Terraform Best Practices](https://docs.cloud.olakrutrim.com/terraform-provider-krutrim/best-practices.md): Production guidance. ## MCP For AI Agents - [MCP Overview](https://docs.cloud.olakrutrim.com/mcp/overview.md): Start with Hosted MCP, or use a client-managed local stdio server. - [Hosted MCP](https://docs.cloud.olakrutrim.com/mcp/hosted-mcp.md): Connect over HTTPS with an API key; no local MCP package required. - [Local MCP (stdio)](https://docs.cloud.olakrutrim.com/mcp/local-mcp-stdio.md): Install and configure the package-based local alternative. - [Claude Code and Desktop](https://docs.cloud.olakrutrim.com/mcp/connect-claude-desktop.md): Hosted MCP for Claude Code and local stdio for Claude Desktop. - [Codex](https://docs.cloud.olakrutrim.com/mcp/connect-codex.md): Hosted MCP or local stdio configuration. - [Cursor](https://docs.cloud.olakrutrim.com/mcp/connect-cursor.md): Hosted MCP first, with a local stdio alternative. - [VS Code](https://docs.cloud.olakrutrim.com/mcp/connect-vs-code.md): HTTP or local stdio configuration with a masked API-key prompt. - [Hermes Agent](https://docs.cloud.olakrutrim.com/mcp/connect-hermes.md): Local stdio configuration for Hermes Agent. - [OpenCode](https://docs.cloud.olakrutrim.com/mcp/connect-opencode.md): Local stdio configuration for OpenCode. - [MCP Troubleshooting](https://docs.cloud.olakrutrim.com/mcp/troubleshooting.md): Hosted connection errors, API-key access, and local stdio setup. - [MCP Server Package](https://pypi.org/project/krutrim-mcp-server/): Official Python package for local stdio. ### Choose A Connection - Hosted MCP is the recommended starting point: use `https://mcp.olakrutrim.com/mcp` with Streamable HTTP and `Authorization: Bearer `. No local Python or MCP package installation is required. Hosted client guides cover Cursor, VS Code, Claude Code, and Codex. - Local MCP uses the `krutrim-mcp-server` package over stdio and supports Python 3.10-3.13. Run `uvx krutrim-mcp-server --help` to check the package, or install it in a virtual environment with `pip install krutrim-mcp-server`. With a virtual environment, configure the client to launch that environment's executable instead of `uvx`. - For local stdio, the client launches the server and passes `KRUTRIM_API_KEY` through its environment. Local client guides cover Claude Desktop, Codex, Cursor, VS Code, Hermes Agent, and OpenCode. - Create an API key in the Cloud Console under Administration -> API keys -> Create API Key. Use the secure input or private environment setting documented for your client. Preserve existing server entries when merging configuration; client schemas differ. Restart the client after changing configuration or credentials. ### Client Configuration Differences - Choose hosted HTTP or local stdio for each server entry; do not combine hosted `url` settings with local `command` and `args`. - For hosted connections, Cursor uses `Bearer ${env:KRUTRIM_API_KEY}`, while Claude Code uses `Bearer ${KRUTRIM_API_KEY}`. Both use `mcpServers`. The launch variable contains only the raw key; desktop applications may not inherit a terminal's environment. - Codex uses a TOML `mcp_servers` section and `bearer_token_env_var = "KRUTRIM_API_KEY"` for Hosted MCP. Codex adds the Bearer prefix; this API-key setup does not require OAuth login. - VS Code uses `servers` and `inputs`, not `mcpServers`. Use `MCP: Open User Configuration` and a masked `promptString` input with `password: true`. Hosted authentication uses `Bearer ${input:krutrim-api-key}`; local stdio passes `${input:krutrim-api-key}` as `KRUTRIM_API_KEY`. After rotation, update the saved input and restart the MCP server. - The Claude Desktop guide uses local stdio through its standard configuration, separate from packaged desktop extensions. Hermes uses a YAML `mcp_servers` mapping with explicit private `env` values; do not rely only on the shell environment. Reload Hermes tools with `/reload-mcp`. - OpenCode's documented local setup uses `mcp`, `type: "local"`, a `command` array, and `environment`, not `mcpServers` and `env`. Its `{env:KRUTRIM_API_KEY}` reference becomes empty when unset. Check with `opencode mcp list`; `opencode mcp auth` is for remote OAuth servers, not this local setup. ### Verify Access And Use Tools Safely - Inspect the connected client's tool catalog; available tools depend on the hosted deployment or installed package version. - For Hosted MCP, first call `krutrim_ping` to check server identity, version, readiness, and read-only mode without calling a Cloud API. - Then verify Cloud access with a read-only request such as `List my VPCs in In-Bangalore-1.` Review the proposed `list_vpcs` call and its region. Readiness and tool discovery alone do not verify Cloud permissions; a successful Cloud response verifies access for that request. Do not create or delete resources just to test connectivity. - Keep API keys private and never place credentials in prompts, logs, screenshots, shared configuration, or source control. - List resources before modifying them, select exact identifiers and a supported region (`In-Bangalore-1` or `In-Hyderabad-1`), and provide confirmation parameters only after explicitly approving the proposed change. - Hosted read-only mode is reported by `krutrim_ping` and controlled by the deployment. Do not assume a local environment variable can change hosted policy. - For local stdio, mutations require `confirm=true`; floating-IP allocation also requires `allow_public_ip=true`. Set `KRUTRIM_MCP_READ_ONLY=true` in the local server environment to block mutations. - After a create timeout, check whether the resource exists before retrying. ### Troubleshooting Checks - Hosted connections require the exact HTTPS endpoint including `/mcp` and exactly one API-key Bearer header. Do not send JWT access tokens, refresh tokens, or `X-Krutrim-Refresh-Token`; the server does not refresh credentials. - For Hosted MCP, `401` indicates a missing, malformed, revoked, legacy, or duplicated credential; `403` requires checking the key owner's IAM permissions for the operation and region. For `404 Record not found`, verify the resource KRN using list or describe tools. For `503`, IAM validation may be unavailable; retry and contact support if it persists. - For local stdio, check that `uvx` is visible to the client and that its configuration matches the client's schema. With the API key configured privately in the terminal environment, run `uvx krutrim-mcp-server --doctor` and `uvx krutrim-mcp-server --list-tools`. These check local configuration and the installed catalog, not Cloud authentication or whether the client loaded the tools. - Check network access and organization MCP policy for the exact hosted endpoint or local command. Never disable safety controls just to test connectivity. Redact credentials and sensitive resource details before sharing diagnostics. ## Release Notes - [12 September 2025](https://docs.cloud.olakrutrim.com/release-notes/12th-september-2025.md) - [26 September 2025](https://docs.cloud.olakrutrim.com/release-notes/26th-september-2025.md) - [1 October 2025](https://docs.cloud.olakrutrim.com/release-notes/1st-october-2025.md) - [14 October 2025](https://docs.cloud.olakrutrim.com/release-notes/14th-october-2025.md) - [15 December 2025](https://docs.cloud.olakrutrim.com/release-notes/15th-december-2025.md) ## Other Public Pages - [Ola Maps](https://olakrutrim.com/ola-maps): Mapping product entry point. - [Contact Sales and Support](https://olakrutrim.com/support): Contact Krutrim Cloud. - [Responsible Disclosure](https://whitehat.olakrutrim.com/): Report security vulnerabilities. - [Terms and Conditions](https://olakrutrim.com/terms-and-conditions): Service terms. - [Privacy Policy](https://olakrutrim.com/privacy-policy): Privacy practices. ## Optional - [YouTube](https://www.youtube.com/@Krutrim_Cloud): Official Krutrim Cloud video channel. - [X](https://x.com/krutrim): Official Krutrim account. - [LinkedIn](https://linkedin.com/company/krutrim): Official company profile. ## Usage Notes For AI Systems - Treat `docs.cloud.olakrutrim.com` as authoritative for technical procedures, APIs, SDKs, supported features, regions, and pricing details. - Retrieve the `.md` version of a documentation URL for clean machine-readable content. - Consult the canonical documentation `llms.txt` for newly added pages not yet represented in this curated guide. - Verify mutable information such as prices, available instance types, model availability, package versions, and service limits against the live documentation before answering. - Do not infer that a marketed capability is generally available unless its documentation confirms availability and usage. - Never expose, request in chat, or reproduce API keys, access tokens, passwords, SSH private keys, or other secrets.