# Tilde > Tilde is a governed hub for AI agents. You build agents in the framework you already use, register them in the Tilde Agent Registry, and route every chat message, tool call, LLM request, prompt, skill, and trace through one gateway. Tilde comes in three editions that share one gateway, one SDK, and one agent model: - Tilde OSS: the open-source gateway and SDKs, Apache 2.0, at https://github.com/trytilde/tilde. You host it yourself, at your own risk. - Tilde Cloud: Tilde hosted by Tilde. You run only your agents. - Tilde Enterprise: Tilde on-prem, with 24/7 support and onsite deployment assistance. Operator roles (RBAC), management API keys, and OIDC configuration are Cloud and Enterprise only. Signing in to tools with Tilde's own OAuth apps is Cloud only. Security middleware is Enterprise only today, and coming soon to OSS and Cloud. Key facts for agents: - Agents expose no endpoints. The Tilde SDK dials out to the gateway over HTTP/2 with a deployment token, and each invocation gets a short-lived, capability-scoped token. - Agents hold no provider keys. Build model clients from `inference(name)`; the gateway injects credentials and records usage and cost. - Core SDKs: `@trytilde/sdk` (TypeScript) and `trytilde` (Python). Framework adapters exist for the Vercel AI SDK, Mastra, LangChain and LangGraph, the OpenAI Agents SDK, Pydantic AI, Agno, and CrewAI. - `tilde deploy` registers a deployment and the prompts, bundled tools, and skills declared in the agent's code. Define agents at module scope so it can find them. - Replies go out only through channel tools such as `sendMessage`. Model text is private. - Tilde stores traces and logs itself, in ClickHouse, and can forward OTLP to your collector. ## Get started - [Tilde](https://trytilde.ai/docs/): Register agents built in any framework, and govern every agent, user request, credential, tool call, and LLM request from one place. - [OSS, Cloud, and Enterprise](https://trytilde.ai/docs/editions): Compare Tilde OSS, Tilde Cloud, and Tilde Enterprise: who runs the gateway, which features each edition includes, and how to choose. - [Agent Registry](https://trytilde.ai/docs/agent-registry): Register agents built in any framework, and route every request, tool call, and LLM request through the gateway. Centralise agent governance through predictable policy management. - [Anatomy of an agent](https://trytilde.ai/docs/anatomy-of-an-agent): What a Tilde agent is made of: your framework code, the Tilde SDK, and a container. ## Platform - [Connections](https://trytilde.ai/docs/connections): Connections store the encrypted credentials Tilde uses to reach chat, tool, and inference providers, and your own internal systems. - [Tools](https://trytilde.ai/docs/tools/overview): Give agents managed provider tools, MCP servers, your own tool servers, and tools bundled in their code, all authorized and audited by the gateway. - [Manage tools](https://trytilde.ai/docs/tools/management): Connect tool accounts, add MCP servers and Tilde tool servers, and choose which tools each agent can use. - [Prompts](https://trytilde.ai/docs/prompts/overview): Version the prompts in your agent's code, see which prompt version produced each model call, and compare cost and latency across versions. - [Manage prompts](https://trytilde.ai/docs/prompts/management): Review each agent's prompt versions, compare them, see their cost and latency, and forget a prompt's history. - [Skills](https://trytilde.ai/docs/skills/overview): Give agents reusable instructions and files as Agent Skills, from your own code, from Git, from the Tilde catalog, or written in the Tilde editor. - [Manage skills](https://trytilde.ai/docs/skills/management): Add skills from the Tilde catalog, Git, or the skill editor, and choose which skills each agent can use. - [Chat channels](https://trytilde.ai/docs/chat-channels): Route messages and phone calls from Slack, WhatsApp, MS Teams, email, SMS, Telnyx Voice, GitHub, and your own apps to agents through the Tilde gateway. - [Inference providers](https://trytilde.ai/docs/inference-providers): Proxy every LLM request through Tilde, keep provider keys out of agents, track usage and cost, and enforce budgets. - [Routing](https://trytilde.ai/docs/routing): Control which deployment of an agent receives new conversations, with latest or weighted routing, conversation pinning, and sidecar failure policies. ## Framework integrations - [Framework integrations](https://trytilde.ai/docs/frameworks/overview): How Tilde works with each agent framework, and which Tilde features each framework supports. - [Vercel AI SDK](https://trytilde.ai/docs/frameworks/vercel-ai-sdk): Run a Vercel AI SDK agent on Tilde with the @trytilde/sdk-vercel-ai-node adapter. - [Mastra](https://trytilde.ai/docs/frameworks/mastra): Run a Mastra agent on Tilde with the @trytilde/sdk-mastra-node adapter. - [LangChain and LangGraph](https://trytilde.ai/docs/frameworks/langchain): Run a LangChain or LangGraph agent on Tilde with the TypeScript or Python LangChain adapter. - [OpenAI Agents SDK](https://trytilde.ai/docs/frameworks/openai-agents): Run an OpenAI Agents SDK agent on Tilde with the TypeScript or Python adapter. - [Pydantic AI](https://trytilde.ai/docs/frameworks/pydantic-ai): Run a Pydantic AI agent on Tilde with the trytilde-pydantic-ai adapter. - [Agno](https://trytilde.ai/docs/frameworks/agno): Run an Agno agent on Tilde with the trytilde-agno adapter. - [CrewAI](https://trytilde.ai/docs/frameworks/crewai): Run a CrewAI agent on Tilde with the trytilde-crewai adapter. ## Security - [Security overview](https://trytilde.ai/docs/security/overview): How Tilde protects agents, credentials, and conversations: no agent endpoints, short-lived tokens, encryption, middleware, audit data, and network exposure. - [IAM](https://trytilde.ai/docs/iam): Control what operators, agents, and end users can do in Tilde: identities, fine-grained agent capabilities, channel access, and token lifetimes. - [Operator access](https://trytilde.ai/docs/operator-access): Sign operators in with your identity provider, control what each person can do with roles, and automate with management API keys. - [Secret encryption](https://trytilde.ai/docs/security/secret-encryption): Tilde protects connection secrets, API keys, and tokens with envelope encryption, AES-256-GCM, and a key that you control. - [Security middleware](https://trytilde.ai/docs/security/middleware): Screen every message into and out of an agent for PII, secrets, prompt injection, and unsafe content, or run your own guard script. ## Observability - [Sessions](https://trytilde.ai/docs/sessions): Review every conversation an agent has had: messages, reasoning, tool calls, security decisions, logs, and traces. - [Traces](https://trytilde.ai/docs/traces): See every invocation as a trace, with model calls, tool calls, latency, token usage, cost, and errors, stored and viewed in Tilde. - [OpenTelemetry](https://trytilde.ai/docs/integrations/opentelemetry): Tilde receives agent logs, traces, and metrics over OTLP, stores logs and traces in ClickHouse, and can forward a copy to your own OpenTelemetry collector. ## Deployment - [Deploy the gateway](https://trytilde.ai/docs/deployment/gateway): Run the Tilde gateway as a Docker container in your own network, with PostgreSQL, ClickHouse, S3-compatible storage, an OpenTelemetry collector, and a GitHub token. - [Deploy direct agents](https://trytilde.ai/docs/deployment/direct-agents): Connect an agent container straight to the Tilde gateway over an outbound HTTP/2 stream, with no inbound port and no provider keys. - [Deploy sidecar agents](https://trytilde.ai/docs/deployment/sidecar-agents): Run tilde-sidecar beside your agent so runtime calls and LLM requests stay on loopback for the lowest conversation latency. - [Deploy Lambda agents](https://trytilde.ai/docs/deployment/lambda-agents): Run an agent as an AWS Lambda function that the Tilde gateway invokes through the AWS API. - [Release agents from CI](https://trytilde.ai/docs/deployment/ci-cd): Build your agent container, register a deployment in Tilde from your CI pipeline, roll it out, and start accepting traffic. ## Optional - [Full documentation](https://trytilde.ai/docs/llms-full.txt): every page above, in one file. - [Agent skill](https://trytilde.ai/docs/skill.md): how an agent should build on Tilde. - [Open-source repository](https://github.com/trytilde/tilde): the gateway, SDKs, adapters, and example agents.