intent driven development

Provider Configuration

Chant supports multiple AI model providers. Configure providers in .chant/config.md to control which model APIs chant uses for spec execution.

Supported Providers

Claude (Default)

Uses the Anthropic Claude CLI (claude command).

Best for:

• Full feature support
• Claude-specific capabilities
• Interactive development

Requirements:

• claude command installed and available in PATH
• Anthropic API key configured

Installation:

pip install anthropic-cli

Configuration:

defaults:
  provider: claude
  model: claude-opus-4-5

No additional endpoint configuration needed.

Ollama

OpenAI-compatible API for running local models.

Best for:

• Local execution
• Offline development
• Cost control
• Privacy-sensitive workloads

Requirements:

• Ollama running locally or accessible via network
• Supported models: Llama, Mistral, and other open-source models

Installation:

# Install Ollama
curl https://ollama.ai/install.sh | sh

# Start Ollama
ollama serve

# Pull a model
ollama pull llama2

Configuration:

defaults:
  provider: ollama
  model: llama2

providers:
  ollama:
    endpoint: http://localhost:11434/v1

Remote Ollama:

providers:
  ollama:
    endpoint: http://ollama-server.example.com:11434/v1

OpenAI

OpenAI API (GPT-4, GPT-3.5, etc.)

Best for:

• Production deployments
• Advanced reasoning tasks
• GPT-specific capabilities

Requirements:

• OpenAI API key (OPENAI_API_KEY environment variable)
• API access

Configuration:

defaults:
  provider: openai
  model: gpt-4

providers:
  openai:
    endpoint: https://api.openai.com/v1

Environment setup:

export OPENAI_API_KEY=sk-...

Azure OpenAI:

defaults:
  provider: openai
  model: gpt-4

providers:
  openai:
    endpoint: https://your-instance.openai.azure.com/openai

Kiro CLI

Kiro CLI (kiro-cli-chat) for MCP-based agent execution.

Best for:

• MCP server integration
• Teams using Kiro ecosystem

Requirements:

• kiro-cli-chat command installed and available in PATH
• MCP servers configured via kiro-cli-chat mcp add

Installation:

# See https://kiro.dev/docs/cli for installation

Configuration:

defaults:
  provider: kirocli
  model: sonnet  # shorthand names: sonnet, opus, haiku

MCP Server Setup: Kiro CLI requires MCP servers to be configured separately:

# Add chant MCP server (command and args must be separate)
kiro-cli-chat mcp add --name chant --command "$(which chant)" --args mcp --scope global

# Verify configuration
kiro-cli-chat mcp list

# Should show under global:
#   • chant        /path/to/chant

Verify tools are available:

# Start interactive chat
kiro-cli-chat chat

# In the chat, type /tools to see available tools
# Should list: chant_spec_list, chant_status, chant_add, etc.

Note: Uses kiro-cli-chat chat --no-interactive --trust-all-tools --model <model> for automated execution.

Configuration Reference

Basic Provider Setup

Set the default provider in .chant/config.md:

# .chant/config.md
---
project:
  name: my-project

defaults:
  provider: claude    # claude | ollama | openai | kirocli
  model: claude-opus-4-5
---

Provider Endpoints

Configure provider-specific endpoints in the providers section:

---
defaults:
  provider: ollama
  model: llama2

providers:
  ollama:
    endpoint: http://localhost:11434/v1
  openai:
    endpoint: https://api.openai.com/v1
---

Model Selection

Specify models for each provider:

Claude:

defaults:
  provider: claude
  model: claude-opus-4-5

Ollama:

defaults:
  provider: ollama
  model: llama2          # or: mistral, mixtral, etc.

OpenAI:

defaults:
  provider: openai
  model: gpt-4           # or: gpt-3.5-turbo, etc.

Split Model Configuration

For chant split operations, specify a separate model:

defaults:
  provider: ollama
  model: llama2
  split_model: mistral   # Used for split operations

If split_model is not specified, it defaults to sonnet (for Claude).

Per-Spec Provider Override

Override the provider for specific specs using frontmatter:

---
status: pending
provider: openai
model: gpt-4
---

# Spec title

This spec will use GPT-4 instead of the default provider.

Global vs. Project Configuration

Global config (~/.config/chant/config.md):

defaults:
  provider: claude
  model: claude-opus-4-5

providers:
  openai:
    endpoint: https://api.openai.com/v1

Project config (.chant/config.md):

project:
  name: my-project

defaults:
  provider: ollama      # Override global default
  model: llama2

Project settings override global settings.

Validation

Verify your provider configuration:

chant config --validate

This checks:

• Provider is supported
• Required fields are present
• Endpoint URLs are valid
• Model names are specified

See Also

• Configuration Reference - Full configuration schema
• CLI Reference - Command reference