Find My AI Tool Logo
All tools

Omnara

Omnara is an open-source API for deploying and running stateful AI agents for users and teams, with support for any model, any machine, and any cloud.

ProductivityDeveloper ToolsAI InfrastructureAI Agents

About Omnara

Omnara

Omnara is presented as an open-source alternative to managed agent platforms such as Claude Managed Agents. It provides one API to deploy and run agents, handling execution and state while the user chooses models, tools, machines, and how users interact with each agent. Agents are described as durable API objects that store state in a durable event log, use no compute while idle, and can be launched, steered, and observed through a REST API, TypeScript SDK, or CLI, with configuration changes applied mid-conversation without redeploying. Omnara offers a hosted cloud (Omnara Cloud) and a self-hosting option under the Apache 2.0 license. The page lists access controls including human approval for risky tool calls, encrypted scoped secrets, role-based access, and an append-only audit history. Models are described as provider-agnostic, including OpenAI, Anthropic, OpenRouter, LiteLLM, and self-hosted models. Integrations include MCP servers and first-party Slack, Discord, and GitHub integrations. The page also lists example agents such as an SRE agent, Slack coding agent, signal digest agent, browser agent, and video generation agent.

Key features

  • Durable agents that store state in a durable event log and recover from failures
  • Agents use no compute while idle
  • Run agents on any machine, including Omnara-provided machines, Blaxel, Daytona, or Unikraft Cloud accounts, or the Omnara daemon on macOS or Linux laptops, VMs, or containers
  • Access controls with human approval for risky tool calls
  • Encrypted, scoped secrets storage
  • Role-based access for people and API keys
  • Append-only audit history of actions
  • REST API, TypeScript SDK, and CLI for launching, steering, and observing agents
  • Change agent behavior mid-conversation by posting a new config without redeploying
  • Model-agnostic support for OpenAI, Anthropic, OpenRouter, LiteLLM, or self-hosted models
  • MCP server and reusable skill connections
  • First-party Slack, Discord, and GitHub integrations
  • Open-source under Apache 2.0 license with self-hosting option
  • Hosted Omnara Cloud with managed infrastructure

Best suited for

  • Teams building internal agents for their organization
  • Product teams adding agents to their own dashboards or applications
  • Engineering teams running agents such as SRE assistants or Slack coding agents
  • Teams wanting to avoid vendor lock-in for models, machines, cloud, and data
  • Organizations that want to self-host agent infrastructure in their own VPC or cloud

Pricing

The platform itself is stated as free. Users can bring their own model keys and machines at no charge, or pay as you go with Omnara credits for models at provider token rates and machines by active time. Self-hosting is stated as free. Startups can apply for Omnara credits.

AI agentsself-hostedMCPagent orchestrationdeveloper toolsopen-sourceagent APImodel-agnostic

Discussion

No comments yet. Start the conversation.