Providers

Bring your own agents

Molly runs the agents and model services you already have set up. It doesn’t ship a model of its own.

  • Codex

    Codex agent

  • Claude

    Claude Agent SDK

  • Cursor

    Cursor Agent

  • Antigravity

    Agent

  • Grok

    Agent

  • Droid

    Factory Droid

  • Kilo

    Kilo Code CLI

  • OpenCode

    OpenCode CLI

  • OpenCode 2

    Separate runtime

  • Pi

    Pi coding agent

Molly agent loop

Molly runs its own tool-using agent loop with a model service you configure. Molly does not provide the underlying model.

Hosted and routed

  • GitHub Copilot
  • OpenRouter
  • DeepSeek
  • Groq
  • Together AI
  • Fireworks AI
  • Cerebras
  • Mistral
  • Hugging Face
  • NVIDIA NIM
  • Moonshot (Kimi)
  • Alibaba Qwen (DashScope)
  • SiliconFlow
  • Gemini API

Cloud endpoints

  • Azure OpenAI
  • Google Vertex AI

Local and self-hosted

  • Ollama
  • LM Studio
  • vLLM
  • LiteLLM

A compatible custom endpoint also works. Local endpoints such as Ollama and LM Studio need an explicit opt-in for plain HTTP.

Before a session starts

A provider listed here isn’t necessarily ready to use. Molly checks each one, in this order.

  1. 01.

    Installed

    Molly finds the provider’s executable or runtime.

  2. 02.

    Enabled

    Your Molly settings allow the provider.

  3. 03.

    Signed in

    The provider reports that you’re authenticated.

  4. 04.

    Ready

    Its current health checks pass.

Capabilities differ

In-session model changes, steering, rollback, skills, slash commands and plugin discovery aren’t available everywhere.

Your data

Your message and the project details for a run go to the agent or model service you choose, under its own privacy terms. Read Molly’s privacy policy.