Skip to content

providers.md

Latest commit

 

History

History
87 lines (60 loc) · 3.18 KB

providers.md

File metadata and controls

87 lines (60 loc) · 3.18 KB

LLM Providers

Agentis supports 12 providers out of the box. Add as many or as few as you want — each is optional.

Configure keys in Settings → Providers.

Supported Providers

Provider Models Key Prefix Get Key
Anthropic Claude Opus 4.6, Sonnet 4.6, Haiku 4.5 sk-ant- console.anthropic.com
OpenAI GPT-4.1, o4-mini, GPT-4o, GPT-4o Mini sk- platform.openai.com
Google AI Gemini 2.5 Pro, Flash, experimental AIzaSy aistudio.google.com
Groq Llama 3.3, Mixtral, Gemma gsk_ console.groq.com
Mistral AI Large, Small, Codestral, Nemo — console.mistral.ai
DeepSeek V3, R1 (reasoning) sk- platform.deepseek.com
OpenRouter 200+ models via one key sk-or- openrouter.ai
Cohere Command R+, Command R — dashboard.cohere.com
xAI (Grok) Grok 3, Grok 3 Mini xai- console.x.ai
Together AI Llama, Mistral, Qwen, DeepSeek — api.together.ai
Ollama Any local model (no key — endpoint) ollama.ai
LM Studio Any GGUF model (no key — endpoint) lmstudio.ai

Recommended Setup

Minimum (free tier works): Anthropic Sonnet 4.6 — best balance of quality and speed.

Best for multi-agent: Add at least 2 providers. Universe uses provider failover — if one is rate-limited, it automatically switches to the next.

Best for cost: Add Groq (free tier, extremely fast) as a secondary provider for lightweight tasks.

Local Providers (No API Key)

Ollama

Run models locally with zero cost.

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull a model
ollama pull llama3.2

# Start server (default: localhost:11434)
ollama serve

In Agentis Settings → Providers → Ollama: set endpoint to http://localhost:11434.

LM Studio

Download LM Studio, load any GGUF model, and start the local server (default port 1234).

In Agentis Settings → Providers → LM Studio: set endpoint to http://localhost:1234.

How Provider Failover Works

In Universe (multi-agent mode), each agent has an assigned provider. If that provider:

  • Returns an error
  • Hits a rate limit
  • Times out

Agentis automatically tries the next available provider in your configured list. This means tasks keep running even if one provider is down.

Cost Reference

Model Input (per 1M tokens) Output (per 1M tokens)
Claude Sonnet 4.6 $3.00 $15.00
Claude Haiku 4.5 $0.80 $4.00
Claude Opus 4.6 $15.00 $75.00
GPT-4o $2.50 $10.00
GPT-4o Mini $0.15 $0.60
Gemini 1.5 Pro $1.25 $5.00
Groq (Llama) Free tier Free tier
Ollama / LM Studio Free (local) Free (local)

Track your actual spend in real time under Analytics.