How to Connect Codex to Datadog (and What It Can't Do)
Yes — Codex can talk to Datadog, and it’s one of the cleaner MCP stories in the observability space. Datadog ships an official, generally available MCP server, and its docs explicitly list OpenAI Codex as a supported client alongside Claude Code and Cursor. Add it to Codex’s MCP configuration and Codex can pull live logs, metrics, traces, and monitor states straight into a debugging session. What it can’t do is operate Datadog for you: it won’t watch monitors overnight, open tickets when something fires, or send your team a weekly error digest. Codex works inside a session you start; the moment the session ends, so does its view of your telemetry.
Here’s what actually works today, how to wire it up, where it stops, and what to use for the always-on layer.
What Codex can do with Datadog
Datadog’s MCP server is a hosted remote server — no local process to babysit — and it went generally available in March 2026. Once Codex is connected, you can ask it to:
- Query logs, metrics, and traces mid-task — “pull the error logs for checkout-service from the last hour and find the regression in this diff.”
- Check monitor and dashboard state — retrieve which monitors are alerting and what the relevant dashboards show, as context for a fix.
- Reach across Datadog products — the server ships 16+ core tools plus optional toolsets for APM, Error Tracking, DBM, Security, and LLM Observability.
- Debug production issues with real telemetry — instead of you copy-pasting stack traces into the prompt, Codex fetches them itself.
This is the genuinely good use case: an AI coding agent that reads your observability data while it writes the fix.
How to set it up
Setup is a developer task, and Datadog’s setup guide covers the specifics:
- Enable the MCP server for your Datadog organization (it’s remote and hosted by Datadog — check the docs for org-level availability and auth).
- Add it to Codex’s config — Codex reads MCP servers from
~/.codex/config.toml(or a project-scoped.codex/config.toml) and supports Streamable HTTP transport, which is what Datadog’s remote server uses. The Codex MCP docs walk through the entry. - Authenticate with your Datadog credentials so the tools are scoped to what your account can see.
Since July 9, 2026, Codex lives as a surface inside the unified ChatGPT desktop app and is available on every plan, metered against your plan’s usage allowance — so there’s no separate Codex subscription to buy, but heavy observability queries burn the same allowance as everything else.
The limits that actually matter
- Session-scoped, not always-on. Codex queries Datadog when you’re in a task. It does not sit there watching monitors at 3am. If a P1 fires while you’re asleep, Codex finds out when you do.
- It reads telemetry; it doesn’t run your incident process. The MCP server is built to query and retrieve observability data. Turning an alert into a Jira ticket, paging the right person, posting the runbook to Slack — that’s workflow territory, not a coding-agent feature.
- Everything assumes a developer at the keyboard.
config.toml, MCP transports, Datadog auth scopes. Fine for engineers; not a connect button an ops lead flips on.
So Codex-plus-Datadog is great for “why is this endpoint slow — check the traces” and not built for “when a monitor fires, run the playbook.”
The always-on layer: Carly
The gap Codex leaves is the operational one — the things that should happen because Datadog said something, whether or not anyone is at a keyboard. That’s what Carly does. Carly is an AI executive assistant that acts on triggers across your stack, set up by conversation instead of config files:
- Tell Carly “when a Datadog monitor fires, open a Jira ticket and post the runbook link to Slack” — in plain English. It interviews you for the details and builds the workflow. No MCP server, no TOML.
- Runs in the cloud, 24/7. Monitors fire at night; Carly acts at night.
- Schedules, too — “every Monday morning, email me a digest of last week’s Datadog alerts grouped by service.”
- Connects the alert to the rest of your work — email (Gmail and Outlook, sent, not just drafted), calendar, CRM, task tools.
Carly integrates with Datadog natively, and connects to 200+ tools — anything without a native connector works through your own API key. AI agents start at $35/month, and steps in a workflow that don’t use AI run free and unlimited.
Codex + Datadog vs Carly
| Codex (Datadog MCP) | Carly | |
|---|---|---|
| Purpose | Pull telemetry into coding sessions | Act on Datadog events as workflows |
| Setup | config.toml + Datadog MCP auth | Describe it in plain English |
| Queries logs/metrics/traces | Yes | Yes |
| Watches monitors 24/7 | No (session-scoped) | Yes |
| Opens tickets, pages, posts to Slack | No | Yes |
| Scheduled digests | No | Yes |
| Built for | Developers debugging code | Teams running operations |
| Pricing | ChatGPT plan allowance | AI agents from $35/mo |
Codex-with-Datadog is a coding agent that can see your telemetry. Carly is an assistant that does something about it.
Frequently Asked Questions
Does OpenAI Codex integrate with Datadog?
Yes, via MCP. Datadog runs an official remote MCP server that explicitly supports OpenAI Codex as a client. Add it to Codex’s config.toml and Codex can query logs, metrics, traces, monitors, and dashboards during a task.
Can Codex alert me when a Datadog monitor fires?
No. Codex only touches Datadog inside a session you’re actively running — it has no background process watching your monitors. For trigger-based actions like “monitor fires → Jira ticket + Slack post,” use a workflow assistant like Carly.
Is the Datadog MCP server official?
Yes. It’s built and hosted by Datadog, went generally available in 2026, and is documented at docs.datadoghq.com/mcp_server. No community shims required.
What does Codex + Datadog cost?
Codex is included in every ChatGPT plan (including Free) since the July 2026 unified app, metered against your plan’s usage allowance. The Datadog MCP server itself is part of your Datadog account; check Datadog’s docs for org availability.
More: Codex + GitHub · Codex + Slack · Codex + Grafana · ChatGPT MCP · Claude + Datadog
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
