masorange · finxo · Apr 14, 2026 · Apr 7, 2026 · Apr 13, 2026 · Apr 13, 2026
diff --git a/.titan/config.toml b/.titan/config.toml
@@ -20,7 +20,4 @@ pr_template_path = ".github/pull_request_template.md"
 auto_assign_prs = true
 
 [plugins.jira]
-enabled = false
-
-[plugins.jira.config]
-default_project = "ECAPP"
+enabled = false
diff --git a/AGENTS.md b/AGENTS.md
@@ -434,8 +434,8 @@ from titan_cli.core.config import TitanConfig
 
 config = TitanConfig()  # Loads and merges global + project
 print(config.config.project.name)
-print(config.config.ai.default)  # Default provider ID
-print(config.config.ai.providers)  # Dict of all configured providers
+print(config.config.ai.default_connection)  # Default AI connection ID
+print(config.config.ai.connections)  # Dict of all configured AI connections
 
 # Check enabled plugins
 if config.is_plugin_enabled("github"):
@@ -566,7 +566,9 @@ class MyCoolPlugin(TitanPlugin):
 
 ## 🤖 AI Integration
 
-Titan CLI includes a modular AI integration layer that allows for interaction with multiple AI providers (Anthropic, Gemini).
+Titan CLI includes a modular AI integration layer that supports multiple AI connections.
+Connections can be either direct providers (Anthropic, OpenAI, Gemini) or LLM gateways
+that expose OpenAI-compatible endpoints such as LiteLLM.
 
 ### File Structure (`ai/`)
 
@@ -580,66 +582,69 @@ titan_cli/ai/
 ├── exceptions.py           # Custom AI-related exceptions
 ├── models.py               # Data models (AIRequest, AIResponse)
 ├── oauth_helper.py         # Helper for Google Cloud OAuth
+├── litellm_client.py       # Shared OpenAI-compatible gateway client
 └── providers/
     ├── __init__.py
     ├── base.py             # AIProvider abstract base class
     ├── anthropic.py
     ├── gemini.py
+    ├── openai.py
+    └── litellm.py
 ```
 
 ### Core Components
 
--   **`AIClient` (`ai/client.py`):** This is the main entry point for using AI functionality. It acts as a facade that reads the user's configuration, retrieves the necessary secrets via `SecretManager`, and instantiates the correct provider. Supports multiple providers with a `provider_id` parameter to select which one to use.
--   **`AIProvider` (`ai/providers/base.py`):** This is an abstract base class that defines the interface for all AI providers. Each provider implements the `generate()` method to interact with its specific API.
+-   **`AIClient` (`ai/client.py`):** Main facade for AI usage. It reads the configured AI connection, retrieves secrets via `SecretManager`, and instantiates the correct direct provider or gateway adapter. Use `connection_id` to select a specific connection.
+-   **`AIProvider` (`ai/providers/base.py`):** Abstract base class implemented by direct providers and the LiteLLM/OpenAI-compatible gateway provider.
+-   **`LiteLLMClient` (`ai/litellm_client.py`):** Shared client for OpenAI-compatible gateways used for connection testing and model discovery.
 
 ### Configuration
 
-AI configuration supports **multiple providers** simultaneously. Users can configure both corporate and personal providers, each with different models and endpoints.
+AI configuration supports **multiple connections** simultaneously.
+Each connection can be:
+- **LLM Gateway**: one endpoint exposing multiple models through an OpenAI-compatible API
+- **Direct Provider**: a direct vendor integration such as Anthropic, OpenAI, or Gemini
 
-AI providers are configured via:
-- Interactive command: `titan ai configure`
-- Main menu option: "AI Configuration" → "Configure AI Provider"
+AI connections are configured from the TUI:
+- Main menu option: `AI Configuration`
+- Then create a new connection and optionally set it as default
 
 The configuration workflow:
-1.  Select configuration type (Corporate or Individual)
-2.  Enter base URL (for corporate endpoints only)
-3.  Select provider (Anthropic, OpenAI, or Gemini)
+1.  Select connection type (`LLM Gateway` or `Direct Provider`)
+2.  For gateways, enter the base URL
+3.  Select the direct provider source when applicable
 4.  Provide API key (stored securely via `SecretManager`)
-5.  Select model with suggestions for popular models
-6.  Assign a friendly name to the provider
+5.  Select or enter the default model
+6.  Assign a friendly name to the connection
 7.  Optionally configure advanced settings (temperature, max_tokens)
-8.  Optionally mark as default provider
+8.  Optionally mark as default connection
 9.  Test the connection
 
-Configuration is stored in the global `~/.titan/config.toml` file with support for multiple providers:
+Configuration is stored in the global `~/.titan/config.toml` file:
 
 ```toml
 [ai]
-default = "corporate-gemini"  # Default provider ID
+default_connection = "work-gateway"
 
-[ai.providers.corporate-gemini]
-name = "Corporate Gemini"
-type = "corporate"
-provider = "gemini"
-model = "gemini-2.0-flash-exp"
+[ai.connections.work-gateway]
+name = "Work Gateway"
+kind = "gateway"
+gateway_type = "openai_compatible"
 base_url = "https://llm.company.com"
+default_model = "gemini-2.5-pro"
 temperature = 0.7
 max_tokens = 4096
 
-[ai.providers.personal-claude]
+[ai.connections.personal-claude]
 name = "Personal Claude"
-type = "individual"
+kind = "direct_provider"
 provider = "anthropic"
-model = "claude-3-5-sonnet-20241022"
+default_model = "claude-sonnet-4-5"
 temperature = 0.7
 max_tokens = 4096
 ```
 
-**Available commands:**
-- `titan ai configure` - Configure a new AI provider
-- `titan ai list` - List all configured providers
-- `titan ai set-default [provider-id]` - Change default provider
-- `titan ai test` - Test connection to default provider
+Titan still loads legacy AI config and migrates it to this structure automatically.
 
 ### Usage
 
@@ -657,15 +662,15 @@ config = TitanConfig()
 secrets = SecretManager()
 
 # 2. Check if AI is configured
-if not config.config.ai or not config.config.ai.providers:
-    print("No AI providers configured. Run: titan ai configure")
+if not config.config.ai or not config.config.ai.connections:
+    print("No AI connections configured.")
     return
 
-# 3. Create the AI client (uses default provider)
+# 3. Create the AI client (uses default connection)
 try:
     ai_client = AIClient(config.config.ai, secrets)
-    # Or specify a specific provider:
-    # ai_client = AIClient(config.config.ai, secrets, provider_id="corporate-gemini")
+    # Or specify a specific connection:
+    # ai_client = AIClient(config.config.ai, secrets, connection_id="work-gateway")
 except AIConfigurationError as e:
     print(f"AI not available: {e}")
     return
@@ -686,12 +691,12 @@ if ai_client.is_available():
     )
     print(creative_response.content)
 
-# 5. Using a specific provider
-corporate_client = AIClient(config.config.ai, secrets, provider_id="corporate-gemini")
-personal_client = AIClient(config.config.ai, secrets, provider_id="personal-claude")
+# 5. Using specific connections
+gateway_client = AIClient(config.config.ai, secrets, connection_id="work-gateway")
+personal_client = AIClient(config.config.ai, secrets, connection_id="personal-claude")
 
-# Each client uses its own provider configuration
-corp_response = corporate_client.generate(messages)
+# Each client uses its own connection configuration
+corp_response = gateway_client.generate(messages)
 personal_response = personal_client.generate(messages)
 ```
 

diff --git a/CLAUDE.md b/CLAUDE.md
@@ -299,8 +299,9 @@ When creating new steps or refactoring existing ones:
 
 - **Python 3.11+**
 - **Textual**: TUI framework
-- **Anthropic SDK**: Claude integration
-- **Google GenAI SDK**: Gemini integration
+- **Anthropic SDK**: Anthropic direct provider
+- **Google GenAI SDK**: Gemini direct provider
+- **OpenAI SDK**: OpenAI direct provider and LiteLLM/OpenAI-compatible gateways
 - **PyGithub**: GitHub API client
 - **Requests**: HTTP client for APIs
 
@@ -340,7 +341,7 @@ in another terminal.
 Titan uses a two-level configuration system:
 
 1. **Global Configuration** (`~/.titan/config.toml`):
-   - AI provider settings (shared across all projects)
+   - AI connection settings (shared across all projects)
    - Global preferences
 
 2. **Project Configuration** (`.titan/config.toml` at the git root):
@@ -356,14 +357,14 @@ Titan uses a two-level configuration system:
 
 Example global config:
 ```toml
-[ai.providers.default]
+[ai]
+default_connection = "default"
+
+[ai.connections.default]
 name = "My Claude"
-type = "individual"
+kind = "direct_provider"
 provider = "anthropic"
-model = "claude-sonnet-4-5"
-
-[ai]
-default = "default"
+default_model = "claude-sonnet-4-5"
 ```
 
 Example project config:
@@ -437,17 +438,19 @@ and run `textual console` in another terminal.
 - ✅ Git plugin (commits, branches, etc.)
 - ✅ GitHub plugin (PRs, issues with AI)
 - ✅ Jira plugin (search, AI-powered analysis)
-- ✅ Claude integration (Anthropic)
-- ✅ Gemini integration (Google)
+- ✅ Anthropic direct provider
+- ✅ OpenAI direct provider
+- ✅ Gemini direct provider
+- ✅ LiteLLM/OpenAI-compatible gateway
 - ✅ All workflow steps migrated to Textual
 
 ### Features
 
 - **Declarative Workflows**: Define flows in YAML
-- **Integrated AI**: Use Claude or Gemini to generate commit messages, PR descriptions, issue analysis
+- **Integrated AI**: Use configured AI connections to generate commit messages, PR descriptions, and issue analysis
 - **Interactive TUI**: Modern interface with Textual
 - **Extensible**: Plugin system
-- **Multi-Provider**: Supports multiple AI providers
+- **Multi-Connection**: Supports direct providers and LLM gateways
 
 ## Recent Important Commits
 
@@ -469,7 +472,7 @@ Titan has been redesigned to work on a per-project basis:
 
 - **Removed**: Global `project_root` and `active_project` settings from `[core]` configuration
 - **New Flow**: Titan must be run from within a project directory
-- **Global Config**: Now only stores AI provider settings (shared across projects)
+- **Global Config**: Now only stores AI connection settings (shared across projects)
 - **Project Config**: Each project has its own `.titan/config.toml` with:
   - Project name
   - Enabled plugins
@@ -499,6 +502,7 @@ Two new wizards guide users through initial setup:
 
 - **Textual Documentation**: https://textual.textualize.io/
 - **Anthropic API**: https://docs.anthropic.com/
+- **OpenAI API**: https://platform.openai.com/docs/
 - **Google GenAI**: https://ai.google.dev/
 
 ---

diff --git a/README.md b/README.md
@@ -9,7 +9,7 @@ Titan CLI is a powerful command-line orchestrator that automates Git, GitHub, JI
 - 🔧 **Project Configuration** - Centralized `.titan/config.toml` for project-specific settings
 - 🔌 **Plugin System** - Extend functionality with Git, GitHub, JIRA, and custom plugins
 - 🎨 **Modern TUI** - Beautiful terminal interface powered by Textual
-- 🤖 **AI Integration** - Optional AI assistance (Claude & Gemini) for commits, PRs, and analysis
+- 🤖 **AI Integration** - Optional AI assistance through direct providers and LLM gateways
 - ⚡ **Workflow Engine** - Compose atomic steps into powerful automated workflows
 - 🔐 **Secure Secrets** - OS keyring integration for API tokens and credentials
 
@@ -59,7 +59,7 @@ titan
 ```
 
 On first run, Titan will guide you through:
-1. **Global Setup** - Configure AI providers (optional)
+1. **Global Setup** - Configure AI connections (optional)
 2. **Project Setup** - Enable plugins and configure project settings
 
 ### Basic Usage
@@ -82,10 +82,10 @@ Titan CLI v1.0.0 includes three core plugins:
 
 ## 🤖 AI Integration
 
-Titan supports multiple AI providers:
+Titan supports multiple AI connections:
 
-- **Anthropic Claude** (Sonnet, Opus, Haiku)
-- **Google Gemini** (Pro, Flash)
+- **Direct Provider**: Anthropic, OpenAI, Gemini
+- **LLM Gateway**: OpenAI-compatible endpoints such as LiteLLM
 
 Configure during first setup or later via the TUI settings.
 

diff --git a/docs/concepts/ai-integration.md b/docs/concepts/ai-integration.md
@@ -1,30 +1,55 @@
 # AI Integration
 
-!!! note "Coming soon"
-    Detailed AI integration documentation is being written.
-
----
-
 ## Overview
 
-Titan supports two AI providers:
+Titan supports AI through configurable **connections**:
 
-- **Anthropic Claude** (Sonnet, Opus, Haiku)
-- **Google Gemini** (Pro, Flash)
+- **Direct Provider** connections for:
+  - Anthropic
+  - OpenAI
+  - Gemini
+- **LLM Gateway** connections for OpenAI-compatible endpoints such as LiteLLM
 
-Configure your provider in `~/.titan/config.toml`:
+Configure AI in `~/.titan/config.toml`:
 
 ```toml
-[ai.providers.default]
-name = "Claude"
-type = "individual"
-provider = "anthropic"
-model = "claude-sonnet-4-5"
-
 [ai]
-default = "default"
+default_connection = "default"
+
+[ai.connections.default]
+name = "My Anthropic"
+kind = "direct_provider"
+provider = "anthropic"
+default_model = "claude-sonnet-4-5"
 ```
 
 Titan stores your API key securely in the OS keyring — you'll be prompted for it on first use.
 
-AI is **optional**. All built-in workflows work without it; AI steps are simply skipped if no provider is configured.
+AI is **optional**. All built-in workflows work without it; AI steps are skipped if no AI connection is configured.
+
+## Connection Types
+
+### Direct Provider
+
+Use this when Titan talks directly to the vendor API.
+
+```toml
+[ai.connections.personal-openai]
+name = "Personal OpenAI"
+kind = "direct_provider"
+provider = "openai"
+default_model = "gpt-5"
+```
+
+### LLM Gateway
+
+Use this when a single endpoint exposes one or more models through an OpenAI-compatible API.
+
+```toml
+[ai.connections.work-gateway]
+name = "Work Gateway"
+kind = "gateway"
+gateway_type = "openai_compatible"
+base_url = "https://llm.company.com"
+default_model = "claude-sonnet-4-5"
+```
diff --git a/docs/getting-started/installation.md b/docs/getting-started/installation.md
@@ -52,7 +52,7 @@ titan
 
 On first launch, two setup wizards run automatically:
 
-1. **Global setup** — Configure AI providers (Claude, Gemini). This is optional and can be skipped.
+1. **Global setup** — Configure AI connections. This is optional and can be skipped.
 2. **Project setup** — Choose a project name and enable the plugins you want (Git, GitHub, Jira).
 
 After setup, the main menu appears and you're ready to run workflows.
@@ -63,7 +63,7 @@ After setup, the main menu appears and you're ready to run workflows.
 
 | Path | Purpose |
 |------|---------|
-| `~/.titan/config.toml` | Global config: AI provider credentials |
+| `~/.titan/config.toml` | Global config: AI connections and credentials |
 | `.titan/config.toml` | Project config: enabled plugins and settings |
 
 The project config lives at the **git repository root**, so it works correctly in monorepos.

diff --git a/docs/getting-started/quick-start.md b/docs/getting-started/quick-start.md
@@ -68,14 +68,14 @@ enabled = false
 Edit `~/.titan/config.toml`:
 
 ```toml
-[ai.providers.default]
-name = "Claude"
-type = "individual"
-provider = "anthropic"
-model = "claude-sonnet-4-5"
-
 [ai]
-default = "default"
+default_connection = "default"
+
+[ai.connections.default]
+name = "My Anthropic"
+kind = "direct_provider"
+provider = "anthropic"
+default_model = "claude-sonnet-4-5"
 ```
 
 Titan will prompt for your API key on first use and store it securely in your OS keyring.