# Obot v0.25.0 > Obot is an open-source AI governance platform with MCP and LLM gateways, MCP hosting, MCP and skills registries, and device management. Use examples and configuration from this version together. This documentation describes release v0.25.0. Prefer individual pages below. Each has version metadata and its HTML source URL. ## Complete version (optional) - [Full documentation](https://docs.obot.ai/v0.25.0/llms-full.txt): 447733 bytes; approximately 111934 tokens (bytes / 4 estimate; tokenizer-dependent). ## Documentation - [Overview](https://docs.obot.ai/v0.25.0/index.md): Obot is an open-source platform for organizations to manage, secure, and govern their AI ecosystems. ## Concepts - [MCP Hosting](https://docs.obot.ai/v0.25.0/concepts/mcp-hosting.md): Obot deploys and manages hosted MCP server workloads on the underlying Kubernetes or Docker runtime. - [MCP Registry](https://docs.obot.ai/v0.25.0/concepts/mcp-registry.md): The MCP Registry is a central place to list and discover MCP servers. It provides a curated collection of servers available to users based on their access permissions. - [MCP Gateway](https://docs.obot.ai/v0.25.0/concepts/mcp-gateway.md): The MCP Gateway runs inside Obot and proxies traffic between MCP clients and hosted or remote MCP servers. - [Obot Agent](https://docs.obot.ai/v0.25.0/concepts/obot-agent.md): Obot Agent is a chat interface built to work directly with MCP. It provides a conversational way for users to interact with MCP servers and accomplish tasks using AI. - [Architecture](https://docs.obot.ai/v0.25.0/concepts/architecture.md): Obot connects AI clients and user devices with hosted services and external providers. It provides MCP and LLM gateways, hosting, identity and access control, audit logs, and MCP and Skills registries. ## Features - [Overview](https://docs.obot.ai/v0.25.0/functionality/overview.md): The MCP Platform is Obot's unified management interface for deploying, managing, and operating MCP servers. It provides role-based access to server management, registries, audit logs, usage tracking, and platform administration. - [MCP Servers](https://docs.obot.ai/v0.25.0/functionality/mcp-servers.md): Managing MCP servers in the MCP Platform - [MCP Tunnels](https://docs.obot.ai/v0.25.0/functionality/mcp-tunnels.md): Connect Obot to remote MCP servers on private networks through an outbound WebSocket tunnel. - [MCP Access Policies](https://docs.obot.ai/v0.25.0/functionality/mcp-access-policies.md): MCP Access Policies control which MCP servers are available to which users. Administrators use access policies to map server entries from the MCP Servers page to specific users and groups, ensuring each team has access to the tools they need. - [MCP Registry API](https://docs.obot.ai/v0.25.0/functionality/mcp-registry-api.md): Obot implements the MCP Registry specification, enabling MCP clients to programmatically discover available servers. - [Audit Logs & Usage](https://docs.obot.ai/v0.25.0/functionality/audit-logs-and-usage.md): The MCP Platform provides visibility into MCP and LLM gateway activity through audit logs and usage tracking. These features help with monitoring, compliance, and understanding how MCP servers and LLM gateway models are being used. - [Filters](https://docs.obot.ai/v0.25.0/functionality/filters.md): Filters are a powerful mechanism for inspecting and controlling tool calls and their results in the MCP Gateway. They provide administrators with the ability to implement custom validation, logging, security checks, or other business logic by intercepting tool requests and responses before they are processed. - [Server Scheduling](https://docs.obot.ai/v0.25.0/functionality/server-scheduling.md): Server Scheduling configures pod scheduling behavior for MCP server deployments in Kubernetes. These settings map directly to Kubernetes Deployment spec fields and control where and how MCP server pods run. - [Obot Agent Management](https://docs.obot.ai/v0.25.0/functionality/obot-agent-management.md): Obot Agent Management provides administrators with tools to configure default agent settings and monitor agent and conversation activity. Access these features from Obot Agent Management in the sidebar. - [Model Access Policies](https://docs.obot.ai/v0.25.0/functionality/model-access-policies.md): Model Access Policies control which users and groups can use which language models in Obot Agent. Administrators create policies to grant model access based on organizational needs—whether that means giving everyone access to standard models, restricting powerful models to specific teams, or anything in between. - [LLM Gateway](https://docs.obot.ai/v0.25.0/functionality/llm-gateway.md): The Obot LLM Gateway lets you call OpenAI, Anthropic, Generic Responses Compatible, Amazon Bedrock, and Azure models through Obot using an Obot API key instead of provider credentials. - [Message Policies](https://docs.obot.ai/v0.25.0/functionality/message-policies.md): Message Policies let administrators enforce content rules written in natural language on Obot Agent traffic at the LLM proxy layer. A policy can apply to: - [Skills](https://docs.obot.ai/v0.25.0/functionality/skills.md): Skills are reusable, structured instructions that agents can discover and install to expand their capabilities. - [Skill Access Policies](https://docs.obot.ai/v0.25.0/functionality/skill-access-policies.md): Skill Access Policies control which users and groups can discover and install which skills. - [Device Management](https://docs.obot.ai/v0.25.0/functionality/device-management.md): Device management gives administrators visibility into the AI clients, MCP servers, skills, and plugins configured on user workstations. It can also audit and control tool calls made by supported local AI clients. - [User Management](https://docs.obot.ai/v0.25.0/functionality/user-management.md): User Management provides administrators with tools to manage users, configure roles, and set up authentication. Access these features from User Management in the sidebar. - [Agent Authorization Scopes](https://docs.obot.ai/v0.25.0/functionality/agent-auth-scopes.md): Agent authorization scopes provide programmatic access to Obot from scripts, automation tools, external MCP clients, and compatible LLM clients. Instead of using interactive browser-based OAuth authentication, you can create scopes that generate API keys with only the capabilities each integration needs. - [Branding](https://docs.obot.ai/v0.25.0/functionality/branding.md): Branding allows administrators to customize the visual appearance of the Obot platform. Access this page from Branding in the sidebar. - [Workflow Sharing](https://docs.obot.ai/v0.25.0/functionality/workflow-sharing.md): Workflow sharing lets users publish a workflow from Obot Agent so it can be discovered and installed by other users as a reusable starting point. ## Installation - [Overview](https://docs.obot.ai/v0.25.0/installation/overview.md): This guide helps you choose the right deployment method for your use case. Before selecting a deployment option, it’s useful to understand the core components that make up an Obot installation: - [Docker Deployment](https://docs.obot.ai/v0.25.0/installation/docker-deployment.md): Deploy Obot using Docker for local development, testing, and proof-of-concept scenarios. - [Kubernetes Deployment](https://docs.obot.ai/v0.25.0/installation/kubernetes-deployment.md): Deploy Obot on Kubernetes for production-grade reliability, scalability, and high availability. - [Persistent Storage in Kubernetes](https://docs.obot.ai/v0.25.0/installation/kubernetes-persistent-storage.md): When Obot deploys MCP servers (including Obot Agent workloads) in Kubernetes, those pods need persistent volumes if you want workspace data to survive pod restarts and rescheduling. - [Obot CLI Setup](https://docs.obot.ai/v0.25.0/installation/cli-setup.md): The obot setup command prepares your local workstation to use an Obot server from the command line and from supported local AI clients. - [Enabling Authentication](https://docs.obot.ai/v0.25.0/installation/enabling-authentication.md): This guide covers the step-by-step process to enable and configure authentication in Obot. Configure either the built-in Local provider or an external identity provider to enable user login. The bootstrap user is not implemented to operate as a regular user. - [Google Cloud GKE](https://docs.obot.ai/v0.25.0/installation/reference-architectures/gcp-gke.md): Deploying Obot to Google Kubernetes Engine follows the standard Helm workflow. As a prerequisite, you'll need the following resources set up in your Google Cloud environment: - [Amazon EKS](https://docs.obot.ai/v0.25.0/installation/reference-architectures/aws-eks.md): Deploying Obot to Amazon Elastic Kubernetes Service follows the standard Helm workflow. As a prerequisite, you'll need the following resources set up in your AWS environment: - [Azure AKS](https://docs.obot.ai/v0.25.0/installation/reference-architectures/azure-aks.md): Deploying Obot to Azure Kubernetes Service follows the standard Helm workflow. As a prerequisite, you'll need the following resources set up in your Azure environment: ## Configuration and Operations - [Auth Providers](https://docs.obot.ai/v0.25.0/configuration/auth-providers.md): Authentication providers allow your Obot installation to authenticate users with the identity provider of your choice. Administrators must configure an authentication provider before users can log in. Only one authentication provider can be configured at a time. - [Model Providers](https://docs.obot.ai/v0.25.0/configuration/model-providers.md): The Model Providers page allows administrators to configure and manage various AI model providers. This guide will walk you through the setup process and explain the available options. - [User Roles](https://docs.obot.ai/v0.25.0/configuration/user-roles.md): Obot uses role-based access control to manage what users can do in the MCP Platform. Each role has different permissions and sees different parts of the interface. - [MCP Server GitOps](https://docs.obot.ai/v0.25.0/configuration/mcp-server-gitops.md): Obot supports managing MCP servers through Git repositories, enabling GitOps workflows. - [MCP Servers in Kubernetes](https://docs.obot.ai/v0.25.0/configuration/mcp-deployments-in-kubernetes.md): This is an overview of how Obot sets up MCP servers in Kubernetes, and how to change some of the configuration values. - [Image Pull Secrets](https://docs.obot.ai/v0.25.0/configuration/image-pull-secrets.md): Obot can attach Kubernetes image pull secrets to MCP server Deployments when MCP servers run in Kubernetes. Use this when MCP server images are stored in a private registry such as GHCR, Docker Hub, a self-hosted registry, or Amazon ECR. - [MCP Server Egress Control](https://docs.obot.ai/v0.25.0/configuration/mcp-server-egress-control.md): MCP server egress control restricts which external domains Kubernetes-hosted MCP servers can reach. It is intended for production deployments where MCP servers may run third-party code and should only be allowed to call known external services. - [Audit Log Exports](https://docs.obot.ai/v0.25.0/configuration/audit-log-export.md): Obot can export MCP and LLM gateway audit logs to various cloud storage providers for long-term retention. This feature supports both one-time exports and scheduled recurring exports. - [MCP Server OAuth Configuration](https://docs.obot.ai/v0.25.0/configuration/mcp-server-oauth-configuration.md): Some remote MCP servers require OAuth authentication with pre-registered client credentials. Unlike servers that support dynamic OAuth registration, these servers need administrators to configure a static set of OAuth credentials (Client ID and Client Secret) that all users share. - [Obot Configuration Reference](https://docs.obot.ai/v0.25.0/configuration/server-configuration.md): The Obot server is configured via environment variables. The following configuration is available: - [Overview](https://docs.obot.ai/v0.25.0/configuration/encryption-providers/overview.md): Obot can encrypt selected sensitive fields in its database and credential store. Application-level encryption is disabled by default (OBOT_SERVER_ENCRYPTION_PROVIDER=none). Configure a provider to enable it. This protects the fields listed below, not entire records, database files, backups, or exported audit logs. - [AWS KMS](https://docs.obot.ai/v0.25.0/configuration/encryption-providers/aws-kms.md): This guide explains how to set up AWS KMS encryption for Obot. - [Azure Key Vault](https://docs.obot.ai/v0.25.0/configuration/encryption-providers/azure-key-vault.md): This guide explains how to set up an Azure VM with a managed identity that can access an Azure Key Vault for encryption operations. - [Google Cloud KMS](https://docs.obot.ai/v0.25.0/configuration/encryption-providers/google-cloud-kms.md): This guide explains how to set up Google Cloud KMS encryption for Obot. - [Custom Encryption Provider](https://docs.obot.ai/v0.25.0/configuration/encryption-providers/custom-provider.md): This guide explains how to set up custom encryption for Obot using a local encryption key. - [Slack MCP Server](https://docs.obot.ai/v0.25.0/configuration/tutorials/slack-mcp-server.md): Slack has an official remote MCP server that Obot can connect to. This guide explains how to set that up. ## Documentation - [Obot Editions](https://docs.obot.ai/v0.25.0/enterprise/overview.md): Obot is available in three editions. They are all delivered in the same container image, and you can upgrade from one to the next in the app. The editions differ in which capabilities are enabled and what usage limits apply. - [FAQ](https://docs.obot.ai/v0.25.0/faq.md): Frequently asked questions about deploying, configuring, and using Obot.