Give Cursor and Codex access to production logs over MCP

To give Cursor or Codex access to your production logs, connect your log tool's MCP server. In Cursor, add its URL to mcp.json and turn it on. In Codex, run codex mcp add with --url, then codex mcp login. Your logs must already be in a tool with a server, such as Fixter (fixter.dev).

Once it is connected, you ask what broke in the same window you code in, and you fix it there. No more copying log lines into chat.

Fixter is ours, and the examples use it. We wrote these steps from Cursor's, OpenAI's and our own docs. We have not run the setup on Cursor or Codex ourselves, so if a label differs on your screen, trust that client's docs. Using Claude Code? See how to connect Claude Code to your logs.

Cursor and Codex MCP setup, side by side

An MCP server is how the agent talks to another tool, here the one that stores your logs. Both clients connect by URL and sign in through the browser.

From Cursor's and OpenAI's docs on November 5, 2026.¹,²,³

Get your production logs into a tool with an MCP server

An agent can only query logs that are stored somewhere. If you have no log tool yet, this is where Fixter fits. Fixter finds and fixes the issues in your app in production, from your coding agent: errors, failing infrastructure, slow requests and downtime. It finds the root cause and fixes it: in your agent over MCP, or as a pull request ready for your review. Nothing to learn, built for small teams.

Paste this into Cursor or Codex in your project:

Read fixter.dev/llms.txt and set up Fixter.

The agent connects the server, sets up standard OpenTelemetry and checks that your data arrives.⁵ You open one link to get your key. Prefer to do it by hand? Set four environment variables on your service. There is no Fixter SDK.⁶

MCP access is on every plan, including the Free plan.⁷ Paid from €20 a month.

How to connect Cursor to your production logs

1. Add the server to mcp.json

Use .cursor/mcp.json for one project, or ~/.cursor/mcp.json for all of them.¹

{
  "mcpServers": {
    "fixter": {
      "url": "https://mcp.fixter.dev"
    }
  }
}

2. Turn it on and sign in

Open Customize in the sidebar and switch on fixter.¹ A browser window opens. Sign in with your Fixter account.⁴

3. Ask something real

Using Fixter, what services logged errors in the last hour?

No tools showing? Open the Output panel and choose MCP Logs.¹

How to connect Codex to your production logs

1. Add the server and sign in

Run this in a terminal. Do not skip the login, it is a separate step.²,³

# in a terminal
codex mcp add fixter --url https://mcp.fixter.dev
codex mcp login fixter

Or add the server to the config file by hand.²

# ~/.codex/config.toml, or .codex/config.toml in a trusted project
[mcp_servers.fixter]
url = "https://mcp.fixter.dev"

The Codex CLI, IDE extension and ChatGPT desktop app share this setup, so you only do it once.²

2. Ask something real

Type /mcp in the terminal UI to see your servers.² Then ask the same first question: "Using Fixter, what services logged errors in the last hour?"

In either client, no browser means no sign-in, for example in CI. Create an API key in Fixter, send it as a bearer token and keep it in your secrets.⁴

Five questions to ask once it is connected

Ask in plain language. The agent picks the tools.⁴

1. What new errors showed up since this morning's deploy?
2. Why did checkout fail last night?
3. What was logged right before this timeout?
4. Which users hit this error since the deploy? (This one needs a user id on your logs.)
5. Show the slowest endpoints today.

The answer comes back in the session you are coding in:

# example output, not from a real account
> what new errors showed up since this morning's deploy?

Two new issues since the 09:40 deploy. A timeout on the payments
service, 142 times, all in the EU region. And 38 failed signups
from a null email field in the signup handler. The first started
two minutes after the deploy.

Then ask the agent to open the code behind the error and propose a fix. It reads your code on your machine. Nothing in this setup sends your repository to Fixter.

What your agent can see and change

Log queries only read. Other Fixter tools can change things, such as alert rules and issue status. Those act right away, with your permissions.⁴ Cursor asks before it uses an MCP tool by default.¹ In Codex, set default_tools_approval_mode = "writes" on the server to get asked before changes.²

Query results reach the model your client uses. If your logs carry personal data, redact it first.

Already on Sentry or Datadog?

If your team is happy there, connect that tool's MCP server. It is less work than moving. Sentry publishes a Cursor entry,⁸ and Datadog lists its setup in its docs.⁹ We did not find a Codex line on either page, so check with the vendor. See 10 monitoring MCP servers, compared.

Fixter fits when you have no monitoring yet, or when you want to be told what broke and why, by email or Slack, rather than go looking. See what Fixter does.

Frequently asked questions

How do I give Cursor access to my production logs?

Add your log tool's MCP server to .cursor/mcp.json or ~/.cursor/mcp.json, for example a server named fixter with the url https://mcp.fixter.dev. Turn it on under Customize in the sidebar and sign in when the browser opens.

How do I give Codex access to my production logs?

Run codex mcp add fixter --url https://mcp.fixter.dev in a terminal, then codex mcp login fixter to sign in through the browser. The Codex CLI, IDE extension and ChatGPT desktop app share that setup.

Can Cursor or Codex read my production logs by default?

No. You have to connect the MCP server of the tool that stores your logs and sign in. Your logs also have to be in that tool first. MCP lets the agent query them; it does not collect them.

Does Fixter get my code when I connect it over MCP?

Not through this setup. MCP carries your production data to the agent, and the agent reads your code on your machine. Monitoring and alerting never touch your code.

Is MCP access on Fixter's free plan?

Yes. MCP access is on every plan, including Free. Paid plans start at €20 a month.

Can I connect Sentry or Datadog to Cursor or Codex instead?

Yes, if your data is already there. Sentry publishes a Cursor entry, and Datadog lists its setup in its docs. Follow the vendor's steps. We did not find a Codex line on either page.