Provider Connect (BYOK)

import { Badge } from ‘@astrojs/starlight/components’;

Agento is designed with a pluggable model provider architecture that lets you Bring Your Own Key (BYOK) or connect to self-hosted LLM endpoints.

Supported Providers

Provider	Recommended Models	Authentication
Anthropic	Claude 3.7 Sonnet, Claude 3.5 Sonnet	ANTHROPIC_API_KEY
OpenAI	GPT-4o, o3-mini	OPENAI_API_KEY
Google Gemini	Gemini 2.0 Flash, Gemini 2.0 Pro	GEMINI_API_KEY
Ollama (Local)	DeepSeek-R1, Qwen 2.5 Coder, Llama 3.3	Local endpoint (http://localhost:11434)
Groq	Llama 3.3 70B, Qwen 2.5 Coder	GROQ_API_KEY
OpenAI-Compatible	Custom vLLM, LiteLLM, or LM Studio	OPENAI_BASE_URL + API Key

---

Configuration Scopes

Provider credentials and model selections can be configured in two scopes:

1. Global Configuration: ~/.config/agento/config.json (applies across all projects on your machine).
2. Project Configuration: <workspace>/.agento/config.json (overrides models for a specific repository).

Example Configuration

{
  "$schema": "https://agento.dev/schemas/config.json",
  "providers": {
    "default": {
      "provider": "anthropic",
      "model": "claude-3-7-sonnet-20250219"
    },
    "planning": {
      "provider": "anthropic",
      "model": "claude-3-7-sonnet-20250219"
    },
    "fastCheck": {
      "provider": "gemini",
      "model": "gemini-2.0-flash"
    }
  }
}

---

Environment Variable Overrides

You can also export standard environment variables in your shell:

export ANTHROPIC_API_KEY="sk-ant-api03-..."
export OPENAI_API_KEY="sk-..."
export GEMINI_API_KEY="AIzaSy..."

---

Model Specialization

Agento allows assigning specialized models to different sub-tasks:

• Planning Model: Deep-reasoning models for architectural planning and plan review.
• Execution Model: Fast, accurate code generation models for Act Mode.
• Micro-Analysis: Ultra-fast lightweight models for rapid diagnostic summarization.