Refactor apps-mcp to use CLI-based approach #4003
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
igrekun
requested review from
a team,
andrewnester,
anton-107,
denik,
lennartkats-db,
pietern and
shreyas-goenka
as code owners
igrekun
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
igrekun
force-pushed
the
mcp/cli-based
branch
from
378d5f3 to
70ee03b
Compare
igrekun
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
igrekun
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
igrekun
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
igrekun
requested review from
fjakobs
and removed request for
andrewnester,
anton-107,
denik,
pietern and
shreyas-goenka
arsenyinfo
approved these changes
Nov 25, 2025
igrekun
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
Collaborator
|
Commit: 7bd295a
19 failing tests:
Top 28 slowest tests (at least 2 minutes):
|
igrekun
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
igrekun
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
igrekun
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
fjakobs
requested changes
Nov 27, 2025
| resp, err := w.StatementExecution.ExecuteAndWait(ctx, dbsql.ExecuteStatementRequest{ | ||
| WarehouseId: warehouseID, | ||
| Statement: statement, | ||
| WaitTimeout: "50s", |
Contributor
There was a problem hiding this comment.
What happens on timeout? Since you're wrapping the statement, the AI might not be able to poll for progress?
|
|
||
| func readClaudeMd(ctx context.Context, configFile string) { | ||
| showFallback := func() { | ||
| cmdio.LogString(ctx, "\nConsult with CLAUDE.md provided in the bundle if present.") |
Contributor
There was a problem hiding this comment.
Note that AGENTS.md is supported more broadly by agents, including those from Cursor, so ideally we support that as well.
Contributor
There was a problem hiding this comment.
we will keep a link in the future. Claude does not read AGENTS.md by default (at least it did not).
| Examples: | ||
| experimental apps-mcp tools init-template # Choose from built-in templates | ||
| experimental apps-mcp tools init-template default-python # Python jobs and notebooks |
Contributor
There was a problem hiding this comment.
I'm not sure I understand why we need this command? Is it adding things that bundle init wasn't doing?
And for AI agents i think we should just always use template-minimal and then add resource-specific things like an app template on top. Ideally those resource-specific additions would be done by the bundle generate command (but we don't have any of that yet so it's good to just start with internal tool commands).
Contributor
Author
There was a problem hiding this comment.
This command mimics the bundle init, but looks for claude.md and on success injects its contents in the output such that the agent definitely gets the instructions. It does a couple more things
- Lets the agent discover the schema of the template to see which parameters it needs
- Allows to pass those parameters as a flag without locking into interactive prompts.
| 1 - Validation failed | ||
| 2 - Invalid flags or configuration`, | ||
| Example: ` databricks experimental apps-mcp tools validate ./my-project | ||
| databricks experimental apps-mcp tools validate ./my-project --validator=custom --custom-command="./validate.sh"`, |
Contributor
There was a problem hiding this comment.
Is --custom-command this actually used? If not maybe we should be agile here?
Some thoughts on how I think a validation tool call might work in the future
- It could be ideal to have a single "validate" that runs all checks regardless of stack like nodejs
- We'd probably want to istinguish between all
fastchecks and allthorough/slow checks. I could imagine that all thefastchecks run all the local tests/lints/etc. Andthoroughmight run slow-running things that depend on compute like the Pipeline "dry run." - Any custom scripts might factor into this via the
scriptssection in databricks.yml. Avalidateorvalidate_fastscript could be executed as part of afastcheck. - Maybe we need an option to select a resource to check (like which pipeline, which app, which endpoint). Rather than a check? If we need that option at all.
Contributor
Author
There was a problem hiding this comment.
Right, it's not used now - serving as a placeholder for future extensions. Ideally we should resolve what validate should do based on the contents of the project folder and have very few flags here.
| @@ -1,18 +1,13 @@ | |||
| Your session in Databricks MCP has been successfully initialized. Here are the guidelines to follow while working on projects using databricks_mcp tools: | |||
| Your session in Databricks MCP has been successfully initialized. | |||
Contributor
There was a problem hiding this comment.
When would this show exactly? It might not still show after context compaction/clearing?
Contributor
Author
There was a problem hiding this comment.
This one is shown on the first tool call regardless of was it explore or invoke_databricks_cli. Compaction and clearing do not reset us on the MCP side since we can't capture those calls from claude / cursor.
Contributor
lennartkats-db
left a comment
lennartkats-db
left a comment
There was a problem hiding this comment.
Good unification/refactoring work, @igrekun! I left some inline comments
## Changes <!-- Brief summary of your changes that is easy to understand --> ## Why <!-- Why are these changes needed? Provide the context that the reviewer might be missing. For example, were there any decisions behind the change that are not reflected in the code itself? --> ## Tests <!-- How have you tested the changes? --> <!-- If your PR needs to be included in the release notes for next release, add a separate entry in NEXT_CHANGELOG.md as part of your PR. -->
igrekun
temporarily deployed
to
test-trigger-is
GitHub Actions
Inactive
fjakobs
approved these changes
Nov 28, 2025
fjakobs
reviewed
Nov 28, 2025
| **Initialize an app from template:** | ||
| ``` | ||
| Add a new API endpoint to fetch customer details | ||
| Initialize a new Streamlit app using the Databricks bundle template |
Contributor
There was a problem hiding this comment.
remove reference to streamlit
fjakobs
reviewed
Nov 28, 2025
| - Gives workflow guidance for Databricks Asset Bundles and Apps | ||
|
|
||
| *This is the foundation of your application - a working, tested template ready for customization.* | ||
| - **`invoke_databricks_cli`** - Execute any Databricks CLI command |
Contributor
There was a problem hiding this comment.
I turned Args into an array of strings because I didn't want to execute it through a shell. We should reflect this here as well
|
|
||
| *This is the foundation of your application - a working, tested template ready for customization.* | ||
| - **`invoke_databricks_cli`** - Execute any Databricks CLI command | ||
| - Run bundle commands: `bundle init`, `bundle validate`, `bundle deploy`, `bundle run` |
Contributor
There was a problem hiding this comment.
Suggested change
| - Run bundle commands: `bundle init`, `bundle validate`, `bundle deploy`, `bundle run` | |
| - Run bundle commands: `["bundle", "init"]`, `["bundle", "validate"]`, `["bundle", "deploy"]`, `["bundle", "run"]` |
Collaborator
|
Commit: 5759c24
24 failing tests:
Top 50 slowest tests (at least 2 minutes):
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This refactors the apps-mcp server to use CLI commands instead of direct API providers, significantly simplifying the architecture and leveraging existing bundle command functionality.
Changes
New CLI-based provider:
Removed providers and templates:
Clean up old development features:
Why
This change reduces code complexity while providing a more maintainable architecture that reuses existing CLI commands rather than duplicating API logic.
Tests