Getting Started with CCO
Install, configure, and run your first task with the Codified Orchestrator.
Installation
Requirements
Python 3.10+
CCO requires Python 3.10 or higher with pip package manager.
Verify your Python version:
python3 --version
Install via pip
Install CCO in development mode with all dependencies:
pip install -e ".[dev]"
Using -e (editable mode) allows you to run cco from anywhere while keeping the package linked to this source directory.
Verify Installation
Confirm CCO is installed correctly:
cco --version
You should see output like:
cco, version 1.0.0
Installation successful! Continue to Configuration to set up your API endpoints.
Configuration
CCO uses a profile-based configuration system. Initialize a profile with your preferred endpoints:
Initialize Configuration
cco config init --profile direct-minimax
The direct-minimax profile configures CCO to use MiniMax API as the primary endpoint. You can create custom profiles for different providers.
Config File Location
CCO stores its configuration at:
~/.codified-orchestrator/config.yaml
Endpoint Architecture
CCO uses a three-tier endpoint model:
Primary Endpoint
Main LLM provider for task execution
The primary endpoint handles most tasks. Configure this based on your preferred model provider (MiniMax, Ollama, etc.).
Fallback Endpoint
Backup provider for reliability
When the primary endpoint is unavailable or returns an error, CCO automatically falls back to this endpoint.
Escalation Endpoint
Human review trigger
For complex or sensitive tasks that exceed the model's capabilities, CCO can escalate to a human reviewer.
View your current configuration:
cco config show
Your First Task
1. Prepare a Repository
You can use an existing repository or create a test repository:
# Create a test repository mkdir my-test-repo && cd my-test-repo git init # Or use an existing repository # cco doctor --repo /path/to/existing/repo
2. Verify Setup with Doctor
Run the doctor command to verify your setup and check the repository:
cco doctor --repo /path/to/your/repo
Doctor checks:
- Configuration validity and endpoint connectivity
- Repository structure and required files
- Memory store status
- Agent availability
For JSON output (useful for automation): cco doctor --repo /path/to/repo --json
3. Plan a Task
Before executing, preview what CCO will do with the plan command:
cco plan --repo /path/to/repo --task "Inspect context loading"
Plan shows:
- Selected agent(s) for the task
- Relevant context that will be loaded
- Expected changes or actions
- Estimated complexity
4. Execute with Guardrails
Execute the task with mutation guardrails enabled:
cco task --repo /path/to/repo --task "Inspect context loading" --guarded
Understanding --guarded Mode
What --guarded Does
Mutation guardrails for safe task execution
The --guarded flag enables safety checks that:
- Pre-flight checks - Validate environment before changes
- Branch isolation - Create a dedicated branch for changes
- Scope limiting - Restrict file modifications to task-relevant paths
- Confirmation prompts - Ask for explicit approval before destructive actions
- Rollback capability - Abort and revert if issues are detected
Always use --guarded for production repositories. Omit it only for exploratory tasks where you want unrestricted access.
Repository Setup
For CCO to work effectively, your repository needs specific files that provide context about your project.
Required Files
AGENTS.md Required
Agent definitions and project-specific instructions
This file defines:
- Available agents and their capabilities
- Agent prompting conventions
- Trigger patterns for auto-selection
- Project-specific agent customizations
Place at repository root: /AGENTS.md
.context/architecture.md Required
System design and component flow
Documents the technical architecture of your project, including:
- Component diagram and relationships
- Data flow patterns
- Key design decisions
- Technology stack overview
Place in: /.context/architecture.md
.context/decisions.md Required
Architectural decision log (ADRs)
Records significant decisions made during development:
- Why certain approaches were chosen
- Alternatives considered and rejected
- Consequences of major changes
- Historical context for legacy code
Place in: /.context/decisions.md
.context/known-issues.md Required
Recurring bugs and mitigations
Documents known problems and workarounds:
- Current bugs and their status
- Known edge cases
- Available workarounds
- Things to avoid
Place in: /.context/known-issues.md
.context/deployment.md Required
Installation and operation notes
Contains deployment and operations information:
- Environment setup requirements
- Configuration parameters
- Deployment procedures
- Operational runbooks
Place in: /.context/deployment.md
Minimal Repository Structure
A minimal CCO-enabled repository looks like:
your-repo/ ├── AGENTS.md # Agent definitions ├── .context/ │ ├── architecture.md # System architecture │ ├── decisions.md # Architectural decisions │ ├── known-issues.md # Known bugs and issues │ └── deployment.md # Deployment guide ├── src/ # Your source code ├── tests/ # Your tests └── README.md # Project documentation
Using the Sample Repository
CCO includes a sample repository for testing:
cco doctor --repo tests/fixtures/sample_repo
You're now ready to use CCO! Continue to the CLI Reference to explore all available commands.