> For the complete documentation index, see [llms.txt](https://karini-ai.gitbook.io/karini-ai-documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://karini-ai.gitbook.io/karini-ai-documentation/introduction.md). # Introduction Agentic AI Foundation to accelerate innovation Karini AI is a no-code agentic AI foundation platform for building, deploying, and governing production-grade AI agents. Teams describe an agent in plain language or assemble it visually, ground it in their own data, deploy it as a copilot or an event-driven workflow, and keep cost, quality, and access under control from day one. The diagram below shows how the platform covers the full [agentic AI lifecycle](https://www.karini.ai/insights/navigating-genaiops-in-enterprises) in one unified framework.
## What the Platform Does ### Build * **Kai, the natural language agent builder:** describe a workflow in plain language and [Kai](/karini-ai-documentation/kai-natural-language-agent-builder.md) designs the canvas, configures prompts and models, validates each node, and deploys it. * **Prompt Playground:** create prompts, agents, and [deep agents](/karini-ai-documentation/prompt-management/deep-agent-beta.md) from templates, attach tools, and compare models side by side for latency, cost, and quality, with a full history of runs. * **Deep agents:** run long, multi-step tasks with human approval points, sub-agents, and persistent state. * **Skills Library:** author, version, and share reusable [agent skills](https://www.karini.ai/insights/skills-library), and pin each agent to the exact version it should run. Kai can turn an SOP into a skill. * **Tools and connectors:** connect agents to databases, ERP and CRM systems, email, SharePoint, web search, custom REST APIs, code interpreters, browser automation, and [MCP servers](/karini-ai-documentation/mcp.md). ### Ground in Your Data * **Knowledge bases:** build [knowledge bases](/karini-ai-documentation/datasets.md) from object stores, enterprise applications, databases, and data warehouses, on vector stores including Amazon OpenSearch, Pinecone, PostgreSQL with pgvector, Azure AI Search, and Teradata EVS. * **Knowledge graphs:** capture relationships between entities with [Graph RAG](/karini-ai-documentation/recipes/graph-rag-recipe.md) on Amazon Neptune. * **Access control:** enforce user, group, and row-level permissions so people only see content they're authorized to view. ### Choose the Right Model * **Model Hub:** use models from Amazon Bedrock, Amazon SageMaker, Databricks, Azure OpenAI, Google Vertex, LiteLLM, and private providers through one [Model Hub](/karini-ai-documentation/model-hub.md), with primary and fallback models for reliability. * **Managed model endpoints:** host embedding and reranker models in Karini, or bring your own fine-tuned models, with automatic scaling down to zero when idle. * **Fine-tuning pipelines:** build dataset-to-training workflows visually with Batch Workflow to distill larger models into smaller, more economical ones. ### Deploy * **Workflow recipes:** assemble agents, routers, guardrails, human-in-the-loop steps, and knowledge bases on a visual canvas with [workflow recipes](/karini-ai-documentation/recipes/workflow-recipe.md), and run them on the Karini runtime or on Amazon Bedrock AgentCore. * **Copilots:** deploy any recipe as a [copilot](/karini-ai-documentation/copilots.md) with custom styling, seed questions, feedback, file upload, and traces. * **Event-driven automation:** trigger recipes from other systems with webhooks, or run them on a schedule. ### Govern and Control Cost * **Guardrails:** block unsafe or out-of-scope requests and responses with [guardrails](/karini-ai-documentation/guardrails.md) at the prompt, recipe, or copilot level. * **Budgets & Controls:** set [daily and monthly budgets](/karini-ai-documentation/budgeting-and-cost-management.md) and rate limits by organization, role, recipe, copilot, and user. * **Prompt caching and semantic caching:** reuse repeated context at a fraction of the cost, and reuse answers to semantically similar requests. ### Observe and Improve * **Observability:** trace every agent step, tool call, and token, and analyze performance and cost trends in built-in dashboards and [observability](/karini-ai-documentation/observability.md) views. * **Evaluation:** score prompts, agents, and workflows against datasets with built-in and custom [evaluation](/karini-ai-documentation/prompt-management/prompt-and-agent-evaluation.md) metrics. * **Continuous improvement:** export conversations to build fine-tuning datasets and improve agents over time. ## Where to Start | If you want to... | Start here | | ------------------------------------- | ------------------------------------------------------------------------ | | Set up your account and resources | [Getting Started](/karini-ai-documentation/getting-started.md) | | Install Karini AI in your environment | [Installation](/karini-ai-documentation/installation.md) | | Build your first agent hands-on | [Agentic AI Workshops](/karini-ai-documentation/agentic-ai-workshops.md) | | Build an agent by describing it | [Kai](/karini-ai-documentation/kai-natural-language-agent-builder.md) |