Files
claude-code-action/docs/configuration.md
Ashwin Bhat e6f32c8321 Remove mcp_config input in favor of --mcp-config in claude_args (#485)
* Remove mcp_config input in favor of --mcp-config in claude_args

BREAKING CHANGE: The mcp_config input has been removed. Users should now use --mcp-config flag in claude_args instead.

This simplifies the action's input surface area and aligns better with the Claude Code CLI interface. Users can still add multiple MCP configurations by using multiple --mcp-config flags.

Migration:
- Before: mcp_config: '{"mcpServers": {...}}'
- After: claude_args: '--mcp-config {"mcpServers": {...}}'

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* Add outer action MCP tests to workflow

- Add test-outer-action-inline-mcp job to test inline MCP config via claude_args
- Add test-outer-action-file-mcp job to test file-based MCP config via claude_args
- Keep base-action tests unchanged (they still use mcp_config parameter)
- Test that MCP tools are properly discovered and can be executed through the outer action

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* Fix: Add Bun setup to outer action MCP test jobs

The test jobs for the outer action were failing because Bun wasn't installed.
Added Setup Bun step to both test-outer-action-inline-mcp and test-outer-action-file-mcp jobs.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* Add id-token permission to outer action MCP test jobs

The outer action needs id-token: write permission for OIDC authentication
when using the GitHub App. Added full permissions block to both
test-outer-action-inline-mcp and test-outer-action-file-mcp jobs.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* Use github_token parameter instead of id-token permission

Replace id-token: write permission with explicit github_token parameter
for both outer action MCP test jobs. This simplifies authentication by
using the provided GitHub token directly.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* Use RUNNER_TEMP environment variable consistently

Changed from GitHub Actions expression syntax to environment variable
for consistency with the rest of the workflow file.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* Use execution_file output from action instead of hardcoded path

Updated outer action test jobs to:
- Add step IDs (claude-inline-test, claude-file-test)
- Use the execution_file output from the action steps
- This is more reliable than hardcoding the output file path

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* tmp

* Fix MCP test assertions to match actual output format

Updated the test assertions to match the actual JSON structure:
- Tool calls are in assistant messages with type='tool_use'
- Tool results are in user messages with type='tool_result'
- The test tool returns 'Test tool response' not 'Hello from test tool'

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* Make inline MCP test actually use the tool instead of just listing

Changed the inline MCP test to:
- Request that Claude uses the test tool (not just list it)
- Add --allowedTools to ensure the tool can be used
- Check that the tool was actually called and returned expected result
- Output the full JSON for debugging

This makes both tests (inline and file-based) consistent in their approach.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

---------

Co-authored-by: Claude <noreply@anthropic.com>
2025-08-26 09:43:18 -07:00

346 lines
11 KiB
Markdown

# Advanced Configuration
## Using Custom MCP Configuration
You can add custom MCP (Model Context Protocol) servers to extend Claude's capabilities using the `--mcp-config` flag in `claude_args`. These servers merge with the built-in GitHub MCP servers.
### Basic Example: Adding a Sequential Thinking Server
```yaml
- uses: anthropics/claude-code-action@v1
with:
anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
claude_args: |
--mcp-config '{"mcpServers": {"sequential-thinking": {"command": "npx", "args": ["-y", "@modelcontextprotocol/server-sequential-thinking"]}}}'
--allowedTools mcp__sequential-thinking__sequentialthinking
# ... other inputs
```
### Passing Secrets to MCP Servers
For MCP servers that require sensitive information like API keys or tokens, you can create a configuration file with GitHub Secrets:
```yaml
- name: Create MCP Config
run: |
cat > /tmp/mcp-config.json << 'EOF'
{
"mcpServers": {
"custom-api-server": {
"command": "npx",
"args": ["-y", "@example/api-server"],
"env": {
"API_KEY": "${{ secrets.CUSTOM_API_KEY }}",
"BASE_URL": "https://api.example.com"
}
}
}
}
EOF
- uses: anthropics/claude-code-action@v1
with:
anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
claude_args: |
--mcp-config /tmp/mcp-config.json
# ... other inputs
```
### Using Python MCP Servers with uv
For Python-based MCP servers managed with `uv`, you need to specify the directory containing your server:
```yaml
- name: Create MCP Config for Python Server
run: |
cat > /tmp/mcp-config.json << 'EOF'
{
"mcpServers": {
"my-python-server": {
"type": "stdio",
"command": "uv",
"args": [
"--directory",
"${{ github.workspace }}/path/to/server/",
"run",
"server_file.py"
]
}
}
}
EOF
- uses: anthropics/claude-code-action@v1
with:
anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
claude_args: |
--mcp-config /tmp/mcp-config.json
--allowedTools my-python-server__<tool_name> # Replace <tool_name> with your server's tool names
# ... other inputs
```
For example, if your Python MCP server is at `mcp_servers/weather.py`, you would use:
```yaml
"args":
["--directory", "${{ github.workspace }}/mcp_servers/", "run", "weather.py"]
```
### Multiple MCP Servers
You can add multiple MCP servers by using multiple `--mcp-config` flags:
```yaml
- uses: anthropics/claude-code-action@v1
with:
anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
claude_args: |
--mcp-config /tmp/config1.json
--mcp-config /tmp/config2.json
--mcp-config '{"mcpServers": {"inline-server": {"command": "npx", "args": ["@example/server"]}}}'
# ... other inputs
```
**Important**:
- Always use GitHub Secrets (`${{ secrets.SECRET_NAME }}`) for sensitive values like API keys, tokens, or passwords. Never hardcode secrets directly in the workflow file.
- Your custom servers will override any built-in servers with the same name.
- The `claude_args` supports multiple `--mcp-config` flags that will be merged together.
## Additional Permissions for CI/CD Integration
The `additional_permissions` input allows Claude to access GitHub Actions workflow information when you grant the necessary permissions. This is particularly useful for analyzing CI/CD failures and debugging workflow issues.
### Enabling GitHub Actions Access
To allow Claude to view workflow run results, job logs, and CI status:
1. **Grant the necessary permission to your GitHub token**:
- When using the default `GITHUB_TOKEN`, add the `actions: read` permission to your workflow:
```yaml
permissions:
contents: write
pull-requests: write
issues: write
actions: read # Add this line
```
2. **Configure the action with additional permissions**:
```yaml
- uses: anthropics/claude-code-action@beta
with:
anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
additional_permissions: |
actions: read
# ... other inputs
```
3. **Claude will automatically get access to CI/CD tools**:
When you enable `actions: read`, Claude can use the following MCP tools:
- `mcp__github_ci__get_ci_status` - View workflow run statuses
- `mcp__github_ci__get_workflow_run_details` - Get detailed workflow information
- `mcp__github_ci__download_job_log` - Download and analyze job logs
### Example: Debugging Failed CI Runs
```yaml
name: Claude CI Helper
on:
issue_comment:
types: [created]
permissions:
contents: write
pull-requests: write
issues: write
actions: read # Required for CI access
jobs:
claude-ci-helper:
runs-on: ubuntu-latest
steps:
- uses: anthropics/claude-code-action@beta
with:
anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
additional_permissions: |
actions: read
# Now Claude can respond to "@claude why did the CI fail?"
```
**Important Notes**:
- The GitHub token must have the `actions: read` permission in your workflow
- If the permission is missing, Claude will warn you and suggest adding it
- Currently, only `actions: read` is supported, but the format allows for future extensions
## Custom Environment Variables
You can pass custom environment variables to Claude Code execution using the `settings` input. This is useful for CI/test setups that require specific environment variables:
```yaml
- uses: anthropics/claude-code-action@v1
with:
settings: |
{
"env": {
"NODE_ENV": "test",
"CI": "true",
"DATABASE_URL": "postgres://test:test@localhost:5432/test_db"
}
}
# ... other inputs
```
These environment variables will be available to Claude Code during execution, allowing it to run tests, build processes, or other commands that depend on specific environment configurations.
## Limiting Conversation Turns
You can limit the number of back-and-forth exchanges Claude can have during task execution using the `claude_args` input. This is useful for:
- Controlling costs by preventing runaway conversations
- Setting time boundaries for automated workflows
- Ensuring predictable behavior in CI/CD pipelines
```yaml
- uses: anthropics/claude-code-action@v1
with:
anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
claude_args: |
--max-turns 5 # Limit to 5 conversation turns
# ... other inputs
```
When the turn limit is reached, Claude will stop execution gracefully. Choose a value that gives Claude enough turns to complete typical tasks while preventing excessive usage.
## Custom Tools
By default, Claude only has access to:
- File operations (reading, committing, editing files, read-only git commands)
- Comment management (creating/updating comments)
- Basic GitHub operations
Claude does **not** have access to execute arbitrary Bash commands by default. If you want Claude to run specific commands (e.g., npm install, npm test), you must explicitly allow them using the `claude_args` configuration:
**Note**: If your repository has a `.mcp.json` file in the root directory, Claude will automatically detect and use the MCP server tools defined there. However, these tools still need to be explicitly allowed.
```yaml
- uses: anthropics/claude-code-action@v1
with:
claude_args: |
--allowedTools "Bash(npm install),Bash(npm run test),Edit,Replace,NotebookEditCell"
--disallowedTools "TaskOutput,KillTask"
# ... other inputs
```
**Note**: The base GitHub tools are always included. Use `--allowedTools` to add additional tools (including specific Bash commands), and `--disallowedTools` to prevent specific tools from being used.
## Custom Model
Specify a Claude model using `claude_args`:
```yaml
- uses: anthropics/claude-code-action@v1
with:
claude_args: |
--model claude-4-0-sonnet-20250805
# ... other inputs
```
For provider-specific models:
```yaml
# AWS Bedrock
- uses: anthropics/claude-code-action@v1
with:
use_bedrock: "true"
claude_args: |
--model anthropic.claude-4-0-sonnet-20250805-v1:0
# ... other inputs
# Google Vertex AI
- uses: anthropics/claude-code-action@v1
with:
use_vertex: "true"
claude_args: |
--model claude-4-0-sonnet@20250805
# ... other inputs
```
## Claude Code Settings
You can provide Claude Code settings to customize behavior such as model selection, environment variables, permissions, and hooks. Settings can be provided either as a JSON string or a path to a settings file.
### Option 1: Settings File
```yaml
- uses: anthropics/claude-code-action@v1
with:
settings: "path/to/settings.json"
# ... other inputs
```
### Option 2: Inline Settings
```yaml
- uses: anthropics/claude-code-action@v1
with:
settings: |
{
"model": "claude-opus-4-1-20250805",
"env": {
"DEBUG": "true",
"API_URL": "https://api.example.com"
},
"permissions": {
"allow": ["Bash", "Read"],
"deny": ["WebFetch"]
},
"hooks": {
"PreToolUse": [{
"matcher": "Bash",
"hooks": [{
"type": "command",
"command": "echo Running bash command..."
}]
}]
}
}
# ... other inputs
```
The settings support all Claude Code settings options including:
- `model`: Override the default model
- `env`: Environment variables for the session
- `permissions`: Tool usage permissions
- `hooks`: Pre/post tool execution hooks
- And more...
For a complete list of available settings and their descriptions, see the [Claude Code settings documentation](https://docs.anthropic.com/en/docs/claude-code/settings).
**Notes**:
- The `enableAllProjectMcpServers` setting is always set to `true` by this action to ensure MCP servers work correctly.
- The `claude_args` input provides direct access to Claude Code CLI arguments and takes precedence over settings.
- We recommend using `claude_args` for simple configurations and `settings` for complex configurations with hooks and environment variables.
## Migration from Deprecated Inputs
Many individual input parameters have been consolidated into `claude_args` or `settings`. Here's how to migrate:
| Old Input | New Approach |
| --------------------- | -------------------------------------------------------- |
| `allowed_tools` | Use `claude_args: "--allowedTools Tool1,Tool2"` |
| `disallowed_tools` | Use `claude_args: "--disallowedTools Tool1,Tool2"` |
| `max_turns` | Use `claude_args: "--max-turns 10"` |
| `model` | Use `claude_args: "--model claude-4-0-sonnet-20250805"` |
| `claude_env` | Use `settings` with `"env"` object |
| `custom_instructions` | Use `claude_args: "--system-prompt 'Your instructions'"` |
| `mcp_config` | Use `claude_args: "--mcp-config '{...}'"` |
| `direct_prompt` | Use `prompt` input instead |
| `override_prompt` | Use `prompt` with GitHub context variables |