Skip to main content
Cursor has native Cekura plugin support. Installing the plugin gives you Cekura Skills and the Cekura MCP server in one step — no separate MCP configuration.

Install the plugin

1

Open the marketplace importer

In Cursor, go to Settings > Plugins > Team Marketplaces > Add Marketplace > Import from Repo.
2

Point it at the Cekura repo

Enter https://github.com/cekura-ai/cekura-skills and import.
3

Install the plugin

Install cekura from the imported marketplace.
4

Authorize MCP

When prompted, complete the OAuth browser sign-in to your Cekura dashboard account. No API key needed — the plugin already includes the MCP server config.
Cursor plugins don’t carry Claude-style slash commands; the bundled Skills cover those workflows and activate automatically when your request matches.

Verify setup

Ask Cursor:
Use the Cekura skills and MCP. List my Cekura agents, pick one, and propose 3 evaluators I should create first. Do not create anything until I approve.
Cursor should list real agents from your workspace and reason in Cekura terms (evaluators, metrics, test profiles, run results).

Keep it updated

Re-import the marketplace from Settings > Plugins > Team Marketplaces to pull the latest version. Auto Refresh (optional) — have Cursor update automatically:
1

Install the Cursor GitHub App

Install it on the cekura-ai/cekura-skills repository (Cursor prompts for this; it’s required for Auto Refresh).
2

Enable Auto Refresh

In Settings > Plugins > Team Marketplaces, find the imported Cekura marketplace and toggle on Auto Refresh.
Cursor then re-indexes the marketplace at most once every ~10 minutes. Auto Refresh updates the existing plugin only — if a brand-new plugin is added to the repo later, re-import the marketplace. Team-wide auto-install (Teams / Enterprise): a workspace admin can mark the Cekura plugin Required for a distribution group in the Team Marketplace settings — that installs and keeps it updated for everyone automatically, with no per-developer action.

No plugin?

Set up Skills + MCP manually, or use the behavior preset fallback.
1

Let Cursor read your agent repo

Ask Cursor to inspect prompts, tool definitions, schemas, and test data so the evaluator plan is grounded in your actual agent behavior.
2

Design coverage before creating anything

Ask for workflow, edge-case, regression, and tool-call coverage. Approve the plan before creating evaluators.
3

Run evaluators through MCP

Once approved, ask Cursor to run the evaluators with a low frequency first. Increase frequency after the first run succeeds.
4

Triage failures in context

Cursor can compare Cekura transcripts against your prompt and codebase, then propose fixes and rerun targeted evaluators.

Troubleshooting

Re-import the marketplace from Settings > Plugins > Team Marketplaces, confirm cekura is installed, and restart Cursor. Make sure you completed the OAuth sign-in when prompted.
Confirm the plugin is installed, then ask explicitly: “Use the Cekura skills to design evaluators.” If you’re on the rules-file fallback instead, confirm .cursor/rules/cekura.md exists.
Sign out of the wrong Cekura account in your browser, re-authorize the plugin, and sign in with the intended account.