Skip to content

[lldb] Document MCP tools & resources #148708

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Jul 16, 2025
Merged

Conversation

JDevlieghere
Copy link
Member

Add documentation for the tools and resources exposed by LLDB's MCP server.

Add documentation for the tools and resources exposed by LLDB's MCP
server.
@llvmbot
Copy link
Member

llvmbot commented Jul 14, 2025

@llvm/pr-subscribers-lldb

Author: Jonas Devlieghere (JDevlieghere)

Changes

Add documentation for the tools and resources exposed by LLDB's MCP server.


Full diff: https://github.com/llvm/llvm-project/pull/148708.diff

1 Files Affected:

  • (modified) lldb/docs/use/mcp.md (+63-1)
diff --git a/lldb/docs/use/mcp.md b/lldb/docs/use/mcp.md
index 375c164fe771c..b7474246b54f3 100644
--- a/lldb/docs/use/mcp.md
+++ b/lldb/docs/use/mcp.md
@@ -75,7 +75,69 @@ Configuration example for [Visual Studio Code](https://code.visualstudio.com/doc
 }
 ```
 
-### Troubleshooting
+## Tools
+
+Tools are a primitive in the Model Context Protocol that enable servers to
+expose functionality to clients.
+
+LLDB's MCP integration exposes one tool, named `lldb_command` which allows the
+model to run the same commands a user would type in the LLDB command
+interpreter. It takes two arguments:
+
+1. The unique debugger ID as a number.
+2. The command and its arguments as a string.
+
+## Resources
+
+Resources are a primitive in the Model Context Protocol that allow servers to
+expose content that can be read by clients.
+
+LLDB's MCP integration exposes a resource for each debugger and target
+instance. Debugger resources are accessible using the following URI:
+
+```
+lldb://debugger/<debugger id>
+```
+
+Example output:
+
+```json
+{
+  "contents": [
+    {
+      "uri": "lldb://debugger/1",
+      "mimeType": "application/json",
+      "text": "{\"debugger_id\":1,\"name\":\"debugger_1\",\"num_targets\":1}"
+    }
+  ]
+}
+```
+
+Debuggers can contain one or more targets, which are accessible using the
+following URI:
+
+```
+lldb://debugger/<debugger id>/target/<target idx>
+```
+
+Example output:
+
+```json
+{
+  "contents": [
+    {
+      "uri": "lldb://debugger/1/target/0",
+      "mimeType": "application/json",
+      "text": "{\"arch\":\"arm64-apple-macosx26.0.0\",\"debugger_id\":1,\"dummy\":false,\"path\":\"/bin/count\",\"platform\":\"host\",\"selected\":true,\"target_idx\":0}"
+    }
+  ]
+}
+```
+
+Note that unlike the debugger id, which is unique, the target index is not
+stable and may be reused when a target is removed and a new target is added.
+
+## Troubleshooting
 
 The MCP server uses the `Host` log channel. You can enable logging with the
 `log enable` command.

model to run the same commands a user would type in the LLDB command
interpreter. It takes two arguments:

1. The unique debugger ID as a number.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this debugger ID mentioned elsewhere, where do you get this ID from?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Oh I see, you read the URI? But you need a debugger ID for that too. Hmm.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All the resources are listed, so you don't need to know it a priori. Depending on the implementation, the model either looks at the resources our you specify one to operate on.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The second bit is what I wondered about. Always annoying to be asked "which X do you want" without knowing how many X there are.

But if the model would present them to you, then that's fine.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In other words -
If I, the human, ever have to get the debugger ID myself, document how I can enumerate them.
If I, the human, don't have to do that, don't change anything.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's the latter, you never specify the debugger ID. At most, you select a resource through the UI, and the model connects the dots behind the scenes that that debugger resource has a certain ID and passes it along here.

Copy link
Collaborator

@DavidSpickett DavidSpickett left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM with whatever is the correct thing to document for debugger IDs.

@DavidSpickett
Copy link
Collaborator

Also, the server made it into 21.x but the docs obviously didn't and there's no release note for it. Is it complete "enough" to cherry-pick this and add a release note?

@JDevlieghere
Copy link
Member Author

Also, the server made it into 21.x but the docs obviously didn't and there's no release note for it. Is it complete "enough" to cherry-pick this and add a release note?

I will add a release note and create a back port request. Do you think it's worth cherrypicking the documentation changes too? I would assume most users consult the website so I usually don't bother, but maybe that assumption isn't accurate.

@DavidSpickett
Copy link
Collaborator

This is unlikely to cause a conflict on the release branch so I'd include it anyway, it's easy to do.

But I agree people are more likely to read the website anyway.

@JDevlieghere JDevlieghere merged commit e8dc96d into llvm:main Jul 16, 2025
12 checks passed
@JDevlieghere JDevlieghere deleted the mcp-docs branch July 16, 2025 16:32
@DavidSpickett
Copy link
Collaborator

Someone running lldb in a secure cluster might be glad to have the page in their install dir :)

tru pushed a commit to llvmbot/llvm-project that referenced this pull request Jul 17, 2025
Add documentation for the tools and resources exposed by LLDB's MCP
server.

(cherry picked from commit e8dc96d)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants