Omnara - Showcase Project

Omnara is an open-source platform for creating and interacting with production-ready AI agents through one API. It handles the infrastructure behind your agents, so you can build personal assistants, internal workflows, and customer-facing applications without developing your own agent control plane. Focus on agents, not infrastructure Taking an agent into production means hosting it, exposing endpoints, storing conversations, recovering from interruptions, and connecting tools and compute. Omnara brings execution, state management, and access to tools and compute together so you can focus on what your agent does. Define instructions, choose a model, and add tools, skills, and MCP integrations. Then create agents, send messages, and interact with them through the API. Bring your own models Use built-in support for OpenAI, OpenRouter, Anthropic, and Amazon Bedrock, with the option to bring your own API keys, or connect other compatible LLM endpoints. Choose models based on the capabilities and cost of each workload rather than defaulting to the most expensive model or committing to one provider. Bring your own computers and sandboxes Connect a local computer, cloud VM, or any environment capable of running Omnara’s lightweight daemon. Give agents a persistent cloud machine or access to files on your laptop while it is online. Use supported sandbox providers through Omnara or bring your own provider API keys. Build around existing resources and choose the environment that fits the task. Reduce idle compute costs Agent history and state persist independently of the execution compute. Durable execution lets agents resume after interruptions, while the serverless architecture reduces the need for a dedicated, always-on server for every agent. Use managed hosting or self-host Use Omnara’s hosted service or deploy the open-source control plane in your own cloud or on premises. Self-hosting gives you control over how agent conversations, logs, and state are stored, accessed, and retained. Choose model endpoints and tools that meet your data-handling requirements. Retained agent history can support evaluations, workflow improvements, and datasets for training or fine-tuning your own models. For developers considering Claude Managed Agents, Omnara offers an open-source, self-hostable alternative with the freedom to choose your models, tools, and execution infrastructure.