Show HN: Omnara – Open-source alternative to Claude Managed Agents
Omnara offers a self-hostable, open-source alternative to proprietary managed AI agent platforms like Claude and OpenAI. Its unique durable architecture, which stores agent state in Postgres and allows agents to span multiple machines, resonates with HN's preference for control and customizability over black-box solutions. This makes it a compelling option for developers seeking flexibility and robustness in their AI agent deployments.
The Lowdown
Omnara is an open-source platform designed to run durable AI agents, presenting itself as a self-hostable alternative to managed services like Claude Managed Agents and OpenAI's Agents API. Built by Kartik from Omnara, this system allows users to deploy agents that manage their state and operations with enhanced flexibility and control.
- Durable Agent Architecture: Omnara agents are created via an API, with their state persisted in Postgres, ensuring they survive restarts and consume no resources when idle. This native durability is a core architectural decision, contrasting with 'bolted-on' durability solutions.
- Stateless Workers & Scalability: The agent loop runs on stateless workers. Each iteration (model call or tool execution) commits in a single transaction, then releases the agent, allowing a single worker to multiplex thousands of agents by picking up the next iteration from a Postgres queue.
- Flexible Machine Integration: Agents can execute commands on any macOS or Linux machine (laptops, cloud sandboxes) by connecting an
omnaraddaemon, eliminating the need for inbound ports. Agents can leverage multiple machines and even transition tasks between an offline laptop and a cloud sandbox. - Comprehensive Feature Set: Omnara supports various model providers (OpenAI, Anthropic, OpenRouter, self-hosted), built-in and custom tools, skills, memory stores, subagents, and schedules. It also offers integrations with platforms like Slack, Discord, and GitHub.
- Control and Observability: The platform includes role-based access control (RBAC), secrets management, permissions for tool execution, streaming events, webhooks, and multi-tenancy. Self-hosted deployments store all agent history in the user's Postgres database, enabling full data control for analytics and training.
- Comparison Advantage: Omnara's key differentiation lies in its open-source nature, self-hostability, model agnosticism, and complete data ownership, offering developers more sovereignty over their AI agent infrastructure compared to proprietary cloud-managed solutions.
By providing an end-to-end, open-source solution for managing and executing AI agents, Omnara empowers developers to build and deploy intelligent systems with unprecedented control over their environment, data, and choice of underlying AI models.
The Gossip
Durable Agents' Destiny
Discussion revolves around the emerging trend of "durable agents" and how Omnara's native durability distinguishes it from other solutions. Commenters note that various platforms are converging on this concept. The author clarifies why Omnara's approach, with durability baked into the harness, is superior to "bolted-on" solutions, which might compromise functionality. This highlights a key architectural advantage and a broader industry shift.
Immediate Implementations
Many users express immediate enthusiasm for Omnara, seeing it as precisely what they needed. Specific use cases like automating persistent keyword watchers across social media are brought up, with the Omnara team providing direct pointers to relevant examples within their documentation. This shows a strong positive reception and practical interest in the project's capabilities.