Om MCP Quickstart
Connect your MCP client to the hosted Om MCP with your Om API key and run your first workflow.
5-step setup
1
Get an API key
Create an Om API key from your dashboard.
export OM_API_KEY="omtx_..."2
Use the hosted MCP URL
Point your MCP client at the hosted Om MCP endpoint.
https://agents.omtx.ai/mcp3
Add client config
In Codex, use `Authorization: Bearer $OM_API_KEY`. In Claude Code and other header-capable clients, send `x-api-key: $OM_API_KEY`.
4
Restart your MCP client
MCP clients need a restart before newly added servers appear in the tool list.
5
Run a smoke test
Ask for `om_status`, `pricing_get`, or `datasets_catalog` before moving into deeper workflows.
Remote connection details
| Field | Value | Why it matters |
|---|---|---|
| Server URL | https://agents.omtx.ai/mcp | This is the hosted MCP endpoint clients connect to. |
| Codex auth | Authorization: Bearer $OM_API_KEY | This is the cleanest supported remote MCP auth path in Codex today. |
| Custom-header auth | x-api-key: $OM_API_KEY | Use this in Claude Code and other clients that support custom headers. |
No local repo setup required
Om MCP is a hosted remote MCP server. You do not need the Om repo, a local Python process, or a manual MCP launcher to use it. You only need the hosted MCP URL and an Om API key.
codex mcp addBash
codex mcp add omtx --url https://agents.omtx.ai/mcp --bearer-token-env-var OM_API_KEYSmoke-test prompts
Run om_status and tell me if the Om MCP is healthy.Run pricing_get and summarize the pricing manifest.Run datasets_catalog and show me the top-level dataset families I can access.Run diligence_search for KRAS and return the highest-signal findings.