# Omnara — MCP server card for AI agents

Deploy durable AI agents with Omnara, an open-source managed-agent platform.

Registry name: `com.omnara/omnara` · category: Cloud and DevOps · version 0.1.0 · updated 2026-09-14

## Connect
- Remote (Claude Code): `claude mcp add --transport http omnara https://app.omnara.com/mcp`

## Facts
- Repository: https://github.com/omnara-ai/omnara (3,417 stars)
- Website: https://www.omnara.com

Before you connect a server: read its repository, and give it only the permissions and keys it needs.

## Links
- AgentGid page: https://agentgid.com/mcp/servers/com-omnara-omnara/


## Omnara: launch and steer cloud AI agents from chat

Omnara is an open-source service that runs AI agents for you on rented or your own computers, around the clock. Its connector lets Claude Code, Codex or Cursor launch those agents, message them, answer their questions and set up their models, machines and schedules. You sign in through the browser, so no key is pasted.

### Where to use it
- Hand a long job to a cloud agent: From your coding assistant, start an Omnara agent that keeps working on its own machine after you close the laptop, for example a code review.
- Answer an agent that waits for you: Agents can stop and ask before a risky step. See what they are waiting on and approve or deny it without opening the dashboard.
- Run an agent on a schedule: Set a daily or weekly time at which an agent gets a message or a new agent starts, such as a morning usage report.
- Set up models, machines and keys: Let the assistant add a model provider, a pool of sandbox machines and stored keys, then share them with one project.
- Give one project's tools to a client: For a client or another tool that should see only some projects, connect it with a limited organization key instead of your own login.

### How to start
- Add the server to your assistant (Claude Code shown; Codex and other clients in the docs).

```
claude mcp add --transport http omnara https://app.omnara.com/mcp
```

- Sign in on the dashboard page the client opens and approve access.
- A new cloud project already has a model provider and a sandbox pool, so you can launch a first agent at once.

### License
The code is open source under Apache 2.0 and free to host yourself. Omnara Cloud has no platform fee, but its models and machines are paid by use.

### Good to know
- The browser sign-in acts as you, with your roles in every organization you belong to, not just one project.
- Agents it launches run commands and edit files on their machines, including your own laptop if you connected it.
- Many tools change or delete things, such as machines, secrets, models and schedules. Keep approving each call.
- Omnara's own models and sandboxes cost money by use; your own model keys and machines are free to connect.
- Keys for scripts can only be created by a person in the dashboard, never by another key.

Checked 2026-11-10 against: https://github.com/omnara-ai/omnara/blob/401a1944bdd5dca2eb7de186ee9a638bf47dcb09/README.md, https://github.com/omnara-ai/omnara/blob/401a1944bdd5dca2eb7de186ee9a638bf47dcb09/server.json, https://github.com/omnara-ai/omnara/blob/401a1944bdd5dca2eb7de186ee9a638bf47dcb09/LICENSE, https://github.com/omnara-ai/omnara/blob/401a1944bdd5dca2eb7de186ee9a638bf47dcb09/internal/httpapi/apimcp/manifest.go, https://github.com/omnara-ai/omnara/blob/401a1944bdd5dca2eb7de186ee9a638bf47dcb09/internal/httpapi/apimcp/server.go, https://github.com/omnara-ai/omnara/blob/401a1944bdd5dca2eb7de186ee9a638bf47dcb09/internal/httpapi/apimcp/filter.go, https://github.com/omnara-ai/omnara/blob/401a1944bdd5dca2eb7de186ee9a638bf47dcb09/internal/httpapi/api_mcp.go, https://docs.omnara.com/api/authentication.md, https://docs.omnara.com/tools/permissions.md, https://docs.omnara.com/llms.txt, https://www.omnara.com/pricing, https://www.omnara.com/

### Mistakes to avoid
- Don't: Giving a shared tool or a client's assistant your own browser sign-in → Do: Create a member-role org API key with roles on only the needed projects and put it in the Authorization header (The browser sign-in carries all your roles in every organization.)
- Don't: Asking the agent to create an API key for itself through the API → Do: A person creates keys in the dashboard (Personal tokens, or API Tokens for org keys) (Key creation, role changes and revoking need a dashboard session.)
- Don't: Committing a config with a pasted omnara_pat_v1_… token → Do: Use the browser sign-in, or keep the token outside the repository (A personal token acts as you with all your access and does not expire on its own.)
- Don't: Expecting a tool that is not listed to work after the connection's roles change → Do: Re-list the tools; the server hides tools your roles do not allow (tools/list is filtered per caller by the operations the caller may run.)

### Questions
- Do I need an API key to connect? No. Clients that support the sign-in flow open the browser; a key is only for clients without it or to limit access to some projects.
- Is it free? Self-hosting is free, and Omnara Cloud has no platform fee. You pay for Omnara's models at provider token rates and for its sandboxes by active time and storage.
- Which models can the agents use? Any model behind an OpenAI- or Anthropic-compatible endpoint, including OpenAI, Anthropic, OpenRouter, Amazon Bedrock, LiteLLM and Ollama.
- Where is the agents' history kept? On Omnara Cloud it is stored by Omnara; when you host it yourself it stays in your own Postgres database.

---
AgentGid's catalog is curated: every entry was checked against the maker's official pages (registry, repository, license and pricing pages) on the date shown; a license value shows the quote it stands on wherever AgentGid has one. Usage numbers come from AgentGid's own scan of public repositories set up for AI coding agents. Claims by makers are reported, not tested by AgentGid.
