feat(claude-code): dynamic model listing and mid-session model switching by codefromthecrypt · Pull Request #7120 · block/goose

codefromthecrypt · 2026-02-10T12:36:05Z

Summary

Before, the claude code provider had a few issues:

subverts the "default" recommendation of claude by hard-coding 'claude-sonnet-4-20250514' which can drift.
does not support model switching mid-session.
does not support listing available models the normal claude /models use.

This changes the provider to get the dynamic list of available model choices from claude, doing no filtering as they are already filtered.
This also implements model switching and adds a helpful log message and test to validate it works.
This also reverse documents that unlisted models are supported, even if there is no list of names.
Finally, this removed buggy ContextLengthExceeded handling as it is not predictable in claude.

Type of Change

Feature
Tests

AI Assistance

This PR was created or reviewed with AI assistance

Testing

New test_model_switch which tests switching from default to sonnet;

Here's a run that also shows model list works, via debug logging:

$ RUST_LOG=debug GOOSE_CLAUDE_CODE_DEBUG=1 cargo test -p goose --test providers -- test_claude_code_provider --nocapture

    Finished `test` profile [unoptimized + debuginfo] target(s) in 0.37s
     Running tests/providers.rs (target/debug/deps/providers-83a00f385f05bffd)

running 1 test
=== claude-code::model_listing ===
[crates/goose/tests/providers.rs:317:9] &models = [
    "default",
    "sonnet",
    "haiku",
]
===================
=== CLAUDE CODE PROVIDER DEBUG ===
Command: "/opt/homebrew/bin/claude"
Original system prompt length: 28 chars
Filtered system prompt length: 28 chars
Filtered system prompt: You are a helpful assistant.
================================
set_model: default resolved to claude-opus-4-6
=== claude-code::basic_response === Hello! 👋 How can I help you today?
=== CLAUDE CODE PROVIDER DEBUG ===
Command: "/opt/homebrew/bin/claude"
Original system prompt length: 28 chars
Filtered system prompt length: 28 chars
Filtered system prompt: You are a helpful assistant.
================================
set_model: sonnet resolved to claude-sonnet-4-5-20250929
=== claude-code::model_switch (default -> sonnet) === Hello! 😊 Hope you're having a great day!
test test_claude_code_provider ... ok

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 15 filtered out; finished in 6.01s


============== Providers ==============
✅ claude-code
=======================================

Related Issues

feat(acp): add model selection support for session/new and session/set_model #7112 ACP model selection
Remove allows_unlisted_models — always allow custom model entry #7103 documents why unlisted models are ok here

Before, the claude provider hard-coded 'claude-sonnet-4-20250514' instead of using "default", didn't support model switching mid-session, and didn't support listing available models dynamically. This uses the stream-json control protocol to fetch models via initialize and switch models via set_model. Also removes buggy ContextLengthExceeded handling as claude-code responses are unpredictable for oversized context. Signed-off-by: Adrian Cole <adrian@tetrate.io>

Copilot

Pull request overview

This PR updates the claude-code provider to rely on Claude CLI model aliases dynamically (instead of a hard-coded dated model), and adds support for switching models during an active provider lifecycle, with accompanying provider integration test updates.

Changes:

Switch claude-code default model to "default" and fetch supported model aliases dynamically via a CLI initialize control request.
Add mid-run model switching via set_model control requests on the persistent CLI subprocess.
Update provider integration tests to cover model switching and adjust context-length testing behavior for claude-code.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 6 comments.

File	Description
crates/goose/src/providers/claude_code.rs	Implements dynamic model listing and adds `set_model` control requests to support model switching on the persistent CLI process.
crates/goose/tests/providers.rs	Extends the provider test harness to optionally test model switching and adjusts context-length testing for `claude-code`.

Copilot · 2026-02-10T12:41:18Z

crates/goose/src/providers/claude_code.rs

+        }
+
+        let _ = child.start_kill();
+        Ok(parse_models_from_lines(&lines))


child.start_kill() sends a kill signal but doesn't await wait() to reap the process; in long-running contexts repeated model listings can leave zombies—prefer child.kill().await followed by child.wait().await (or wait_with_output).

Copilot · 2026-02-10T12:41:19Z

crates/goose/tests/providers.rs

+            .await?;
+
+        assert!(
+            matches!(response.content[0], MessageContent::Text(_)),


This test can panic if a provider returns an empty content array; use a non-panicking check (e.g., response.content.first()) before matching on the item.

Suggested change

matches!(response.content[0], MessageContent::Text(_)),

matches!(response.content.first(), Some(MessageContent::Text(_))),

crates/goose/tests/providers.rs

Copilot · 2026-02-10T12:41:19Z

crates/goose/src/providers/claude_code.rs

+            "request_id": request_id,
+            "request": {"subtype": "set_model", "model": model}
+        });
+        let mut req_str = serde_json::to_string(&req).unwrap();


serde_json::to_string(&req).unwrap() can panic in production; return a ProviderError::RequestFailed (or similar) with the serialization error instead of unwrapping.

Suggested change

let mut req_str = serde_json::to_string(&req).unwrap();

let mut req_str = serde_json::to_string(&req).map_err(|e| {

ProviderError::RequestFailed(format!("Failed to serialize set_model request: {e}"))

})?;

Copilot · 2026-02-10T12:41:20Z

crates/goose/src/providers/claude_code.rs

+        // Read lines until we get the control_response for our request.
+        let mut line = String::new();
+        loop {
+            line.clear();
+            match self.reader.read_line(&mut line).await {
+                Ok(0) => {
+                    return Err(ProviderError::RequestFailed(
+                        "CLI process terminated while waiting for set_model response".to_string(),
+                    ));
+                }
+                Ok(_) => {
+                    let trimmed = line.trim();
+                    if trimmed.is_empty() {
+                        continue;
+                    }
+                    if let Ok(parsed) = serde_json::from_str::<Value>(trimmed) {
+                        if parsed.get("type").and_then(|t| t.as_str()) == Some("control_response") {
+                            let success =
+                                parsed.pointer("/response/subtype").and_then(|s| s.as_str())
+                                    == Some("success");
+                            if success {
+                                self.current_model = model.to_string();
+                                self.log_model_update = true;
+                                return Ok(());
+                            } else {
+                                let err = parsed
+                                    .pointer("/response/error")
+                                    .and_then(|e| e.as_str())
+                                    .unwrap_or("unknown");
+                                return Err(ProviderError::RequestFailed(format!(
+                                    "set_model failed: {err}"
+                                )));
+                            }
+                        }


send_set_model treats the first control_response it sees as the response to the set_model request, but it never verifies request_id matches the one it sent, so an unrelated control_response could incorrectly update current_model and desynchronize the stream; parse and match request_id (and ideally ignore/queue other messages).

Copilot · 2026-02-10T12:41:20Z

crates/goose/src/providers/claude_code.rs

+            "request_id": "model_list",
+            "request": {"subtype": "initialize"}
+        });
+        let mut request_str = serde_json::to_string(&request).unwrap();


In fetch_supported_models, serde_json::to_string(&request).unwrap() can panic; convert this to a fallible serialization with proper ProviderError propagation.

Suggested change

let mut request_str = serde_json::to_string(&request).unwrap();

let mut request_str = serde_json::to_string(&request).map_err(|e| {

ProviderError::RequestFailed(format!("Failed to serialize initialize request: {e}"))

})?;

codefromthecrypt · 2026-02-10T22:39:31Z

finally, a few decent review comments, @copilot!

Copilot · 2026-02-10T22:39:41Z

@codefromthecrypt I've opened a new pull request, #7129, to work on those changes. Once the pull request is ready, I'll request review from you.

michaelneale

nice catch

* origin/main: (107 commits) feat: Allow overriding default bat themes using environment variables (#7140) Make the system prompt smaller (#6991) Pre release script (#7145) Spelling (#7137) feat(mcp): upgrade rmcp to 0.15.0 and advertise MCP Apps UI extension capability (#6927) fix: ensure assistant messages with tool_calls include content field (#7076) fix(canonical): handle gcp_vertex_ai model mapping correctly (#6836) Group dependencies in root Cargo.toml (#6948) refactor: updated elevenLabs API module and `remove button` UX (#6781) fix: we were missing content from langfuse traces (#7135) docs: update username in authors.yml (#7132) fix extension selector syncing issues (#7133) fix(acp): per-session Agent for model isolation and load_session restore (#7115) fix(claude-code): defensive coding improvements for model switching (#7131) feat(claude-code): dynamic model listing and mid-session model switching (#7120) Inline worklet source (#7128) [docs] One shot prompting is dead - Blog Post (#7113) fix: correct spelling of Debbie O'Brien's name in authors.yml (#7127) docs: GCP Vertex AI org policy filtering & update OnboardingProviderSetup component (#7125) feat: replace subagent and skills with unified summon extension (#6964) ... # Conflicts: # Cargo.lock # Cargo.toml

* upstream/main: (109 commits) [docs] Skills Marketplace UI Improvements (block#7158) More no-window flags (block#7122) feat: Allow overriding default bat themes using environment variables (block#7140) Make the system prompt smaller (block#6991) Pre release script (block#7145) Spelling (block#7137) feat(mcp): upgrade rmcp to 0.15.0 and advertise MCP Apps UI extension capability (block#6927) fix: ensure assistant messages with tool_calls include content field (block#7076) fix(canonical): handle gcp_vertex_ai model mapping correctly (block#6836) Group dependencies in root Cargo.toml (block#6948) refactor: updated elevenLabs API module and `remove button` UX (block#6781) fix: we were missing content from langfuse traces (block#7135) docs: update username in authors.yml (block#7132) fix extension selector syncing issues (block#7133) fix(acp): per-session Agent for model isolation and load_session restore (block#7115) fix(claude-code): defensive coding improvements for model switching (block#7131) feat(claude-code): dynamic model listing and mid-session model switching (block#7120) Inline worklet source (block#7128) [docs] One shot prompting is dead - Blog Post (block#7113) fix: correct spelling of Debbie O'Brien's name in authors.yml (block#7127) ...

…ing (block#7120) Signed-off-by: Adrian Cole <adrian@tetrate.io>

codefromthecrypt requested review from DOsinga, alexhancock, Copilot and michaelneale February 10, 2026 12:36

Copilot started reviewing on behalf of codefromthecrypt February 10, 2026 12:36 View session

This was referenced Feb 10, 2026

Remove allows_unlisted_models — always allow custom model entry #7103

Open

fix: allow unlisted models for CLI providers #7090

Merged

codefromthecrypt enabled auto-merge February 10, 2026 12:39

Copilot AI reviewed Feb 10, 2026

View reviewed changes

Copilot AI mentioned this pull request Feb 10, 2026

[WIP] Fix dynamic model listing and mid-session model switching #7129

Closed

michaelneale approved these changes Feb 10, 2026

View reviewed changes

codefromthecrypt added this pull request to the merge queue Feb 10, 2026

Merged via the queue into main with commit 01d425b Feb 10, 2026
24 checks passed

codefromthecrypt deleted the claude-model branch February 10, 2026 22:53

codefromthecrypt mentioned this pull request Feb 10, 2026

fix(claude-code): defensive coding improvements for model switching #7131

Merged

3 tasks

Tyler-Hardin pushed a commit to Tyler-Hardin/goose that referenced this pull request Feb 11, 2026

feat(claude-code): dynamic model listing and mid-session model switch…

f227f54

…ing (block#7120) Signed-off-by: Adrian Cole <adrian@tetrate.io>

Tyler-Hardin pushed a commit to Tyler-Hardin/goose that referenced this pull request Feb 11, 2026

feat(claude-code): dynamic model listing and mid-session model switch…

a3f32ac

…ing (block#7120) Signed-off-by: Adrian Cole <adrian@tetrate.io>

Tyler-Hardin pushed a commit to Tyler-Hardin/goose that referenced this pull request Feb 11, 2026

feat(claude-code): dynamic model listing and mid-session model switch…

1d8db93

…ing (block#7120) Signed-off-by: Adrian Cole <adrian@tetrate.io>

github-actions bot mentioned this pull request Feb 12, 2026

chore(release): release version 1.24.0 (minor) #7102

Open

BrewTestBot mentioned this pull request Feb 12, 2026

block-goose-cli 1.24.0 Homebrew/homebrew-core#267289

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(claude-code): dynamic model listing and mid-session model switching#7120

feat(claude-code): dynamic model listing and mid-session model switching#7120
codefromthecrypt merged 1 commit intomainfrom
claude-model

codefromthecrypt commented Feb 10, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Feb 10, 2026

Uh oh!

Copilot AI Feb 10, 2026

Uh oh!

Uh oh!

Copilot AI Feb 10, 2026

Uh oh!

Copilot AI Feb 10, 2026

Uh oh!

Copilot AI Feb 10, 2026

Uh oh!

codefromthecrypt commented Feb 10, 2026

Uh oh!

Copilot AI commented Feb 10, 2026

Uh oh!

michaelneale left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

	matches!(response.content[0], MessageContent::Text(_)),
	matches!(response.content.first(), Some(MessageContent::Text(_))),

-        let mut req_str = serde_json::to_string(&req).unwrap();
+        let mut req_str = serde_json::to_string(&req).map_err(|e| {
+            ProviderError::RequestFailed(format!("Failed to serialize set_model request: {e}"))
+        })?;

-        let mut request_str = serde_json::to_string(&request).unwrap();
+        let mut request_str = serde_json::to_string(&request).map_err(|e| {
+            ProviderError::RequestFailed(format!("Failed to serialize initialize request: {e}"))
+        })?;

Conversation

codefromthecrypt commented Feb 10, 2026

Summary

Type of Change

AI Assistance

Testing

Related Issues

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Copilot AI Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

codefromthecrypt commented Feb 10, 2026

Uh oh!

Copilot AI commented Feb 10, 2026

Uh oh!

michaelneale left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants