Source: github.com/omnara-ai/omnara
Omnara lets developers run AI agents through a REST API, a TypeScript SDK and a CLI, with tools, secrets, permissions and streaming built in. The project can be self-hosted with Docker Compose, or used through the hosted Omnara Cloud service.
Models can come from OpenAI, Anthropic, OpenRouter, Amazon Bedrock, LiteLLM, Ollama or any compatible endpoint. Agents can run in sandboxes from eleven providers or on a macOS or Linux machine, where a daemon connects outward so no inbound ports are needed. Every step is saved in Postgres, so agents can resume after crashes or restarts.
Agents can be started from Slack, Discord, GitHub, cron schedules or the API. Each tool can run immediately, ask a person first, or be blocked. Omnara also provides an MCP server at app.omnara.com/mcp that Claude Code, Codex and Cursor can use to launch and manage agents.
Why it matters if you build agents
Teams that want to choose their own models, keep agent history in their own Postgres, or avoid a single vendor now have an Apache 2.0 option for hosted-style agents.
Per-tool permission settings and saved state matter for long-running agents that touch laptops or sandboxes.
The project says its local defaults are intentionally insecure, so production deployments should follow its self-hosting guide first.
Get stories like this every weekday
One short email with the day's agent tooling news.