Repolode
Submit
omnara-ai

omnara-ai/omnara

Self-hostable open-source platform for running AI agents on any model and any machine, with tools, permissions and streaming built in.

Open-source alternative to:Claude Managed AgentsOpenAI Agents API

View on GitHubWebsite
Stars
2,947
Forks
237
Open issues
25
Last commit
11 hours ago

What is it?

Omnara is a managed agent platform: you launch AI agents through an API, CLI or TypeScript SDK, and the platform runs each agent's loop, calls the configured model, and executes tools on machines you choose. It is positioned as an open-source, self-hostable alternative to Claude Managed Agents and OpenAI's Agents API, and works with OpenAI, Anthropic, OpenRouter, Amazon Bedrock, LiteLLM, Ollama and other compatible endpoints.

Agent state is saved step by step in Postgres, so agents can resume after crashes, restarts or machine disconnects. Agents can run commands and edit files on a laptop, VM or container through a lightweight daemon that connects outward, or on sandboxes from several providers. Permissions decide whether each tool runs immediately, asks a person, or is blocked.

Integrations include Slack, Discord and GitHub, where agents can start from mentions or pull requests, plus cron triggers, webhooks, MCP servers and custom tools. Self-hosting uses Docker Compose, and Omnara Cloud is the hosted version with the same API and config format.

Key features

  • 01Runs agents on any compatible model, using your own keys or self-hosted models.
  • 02Built-in tools for shell, files, web search and asking a person, plus MCP and custom tools.
  • 03Durable state saves every step so agents resume after crashes or restarts.
  • 04Machines can be local computers, VMs or containers via a daemon, or sandboxes from several providers.
  • 05Per-tool permissions require approval before sensitive actions run.
  • 06Slack, Discord and GitHub integrations, plus cron triggers and webhooks.

Use cases

  • Running a coding agent that fixes issues mentioned in Slack or reviews GitHub pull requests.
  • Hosting internal agents on your own infrastructure while keeping agent history in your own Postgres.
  • Scheduling recurring analysis agents, such as daily product usage reports.
  • Letting one agent work across several machines, such as a set of GPU servers.

Quick start

  1. 1

    Sign in at app.omnara.com to get a project with a model and sandboxes

  2. 2

    Save the sample config as agent.yaml

  3. 3

    Launch the agent with the Omnara CLI (Node.js 22 or newer)

    npx omnara agents launch --file agent.yaml --name "Machine assistant" --message "Report this machine's operating system and free disk space."

Things to know

  • The local self-hosting defaults are intentionally insecure and need hardening before real deployment.
  • A new self-hosted instance has no models or machines, so you must add your own provider keys and machines.

Topics

#ai-agents#agent-platform#llm#self-hosted#mcp#go#typescript#workflow-automation#agent-infrastructure#agent-runtime

FAQ

What is omnara?
Self-hostable open-source platform for running AI agents on any model and any machine, with tools, permissions and streaming built in. Omnara is a managed agent platform: you launch AI agents through an API, CLI or TypeScript SDK, and the platform runs each agent's loop, calls the configured model, and executes tools on machines you choose. It is positioned as an open-source, self-hostable alternative to Claude Managed Agents and OpenAI's Agents API, and works with OpenAI, Anthropic, OpenRouter, Amazon Bedrock, LiteLLM, Ollama and other compatible endpoints.
Is omnara free to use? What license does it have?
omnara is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is omnara still maintained?
Actively maintained. The last commit to omnara was on Oct 20, 2026.
What can I use omnara for?
Running a coding agent that fixes issues mentioned in Slack or reviews GitHub pull requests. Hosting internal agents on your own infrastructure while keeping agent history in your own Postgres. Scheduling recurring analysis agents, such as daily product usage reports. Letting one agent work across several machines, such as a set of GPU servers.
How do I install omnara?
Sign in at app.omnara.com to get a project with a model and sandboxes Save the sample config as agent.yaml Launch the agent with the Omnara CLI (Node.js 22 or newer): npx omnara agents launch --file agent.yaml --name "Machine assistant" --message "Report this machine's operating system and free disk space."
Where can I get omnara?
The source code and installation instructions are on GitHub: https://github.com/omnara-ai/omnara

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects