blob: 28a97e6ebde08a45a7b834033da519439e1cd857 [file] [log] [blame] [view]
# Best Practices
Here are some best practices and tips for using the Gerrit MCP server effectively.
## Use Natural Language
The server's tools are designed to be called by a language model. You don't need to remember the exact tool names or parameters. Just state what you want to do in plain English.
* **Instead of:** `gerrit.query_changes(query="owner:me status:open")`
* **Prefer:** "Show me my open CLs"
* **Instead of:** `gerrit.get_change_details(change_id="12345")`
* **Prefer:** "What are the details for CL 12345?"
## Be Specific to Reduce Noise
Gerrit repositories can be very busy. The more specific your query, the more relevant the results will be.
* **Good:** "Find CLs in the 'fuchsia' project"
* **Better:** "Search for open CLs in the 'fuchsia' project with the word 'refactor'"
* **Best:** "Show me open CLs by `user@example.com` in the `zircon` project from the last week"
## Chaining Commands
You can ask the model to perform a series of actions.
* "Find the most recent CL by `user@example.com`."
* (After the result is returned) "Now, list the files in that CL."
* (After the file list is returned) "Show me the diff for `src/main.py`."
## Use Different Gerrit Instances
If your `gerrit_config.json` is configured with multiple hosts, you can specify which one to use in your prompt.
* "On the **AOSP gerrit**, find CLs related to 'kernel'."
* "Search for CLs on the **internal server** by `user@google.com`."
If you don't specify a host, the `default_gerrit_base_url` from your configuration will be used.