Cline
An open-source agent in your editor where you control the model and pay only for inference.
verified 2026-11-03
Agent harnesses (open source)
Open-source scaffolding that turns an LLM into a working agent: the loop, tools, memory and file access. Bring your own model, from a cloud API to a local one through Ollama or LM Studio.
| Harness | License | Local / offline | Setup |
|---|---|---|---|
| Cline | Apache-2.0 (the JetBrains plugin isn't open source) | yes Docs cover Ollama, LM Studio and Atomic Chat, and recommend turning on "Use Compact Prompt" for local models. | low Install the VS Code extension from the Marketplace, or use npm i -g cline for the CLI, or the macOS/Windows desktop app. |
| Aider | Apache-2.0 | yes Docs cover Ollama (aider --model ollama_chat/<model>, which also sets a large enough context window automatically), LM Studio (lm_studio/<model>) and any local OpenAI-compatible server. | low python -m pip install aider-install && aider-install, or the one-line curl or PowerShell installer, uv, pipx or Docker. |
| OpenCode | MIT | yes Docs show config for Ollama (http://localhost:11434/v1), LM Studio (http://127.0.0.1:1234/v1) and llama.cpp's llama-server through an OpenAI-compatible provider. | low One-line curl installer, npm i -g @opencode/cli, Homebrew, pacman, mise or Nix. |
| Goose | Apache-2.0 | yes Docs say it can "work entirely offline" with a local LLM. | low Desktop app for macOS, Linux and Windows, or the CLI via a one-line install script or Homebrew (brew install block-goose-cli). |
| Hermes Agent | MIT | yes Docs cover fully local setups with Ollama, vLLM and llama.cpp, plus LM Studio as a built-in provider, all through an OpenAI-compatible endpoint (for example http://localhost:11434/v1 for Ollama). | low One-line installer (curl … | bash on Linux, macOS and WSL2; PowerShell on Windows), then hermes setup or hermes model to pick a provider. |
| OpenHands | MIT | yes Docs cover local model servers (LM Studio, Ollama, vLLM and SGLang), and Agent Canvas and its backend run on your own machine by default. | medium npm install -g @openhands/agent-canvas (needs Node.js 24+ and uv), or one docker run of the agent-canvas image for a sandboxed setup. |
Aider
Terminal pair programming with automatic git commits and any LLM, including local models.
verified 2026-11-03
OpenCode
Trying a capable open-source terminal agent for $0, with free models and no account.
verified 2026-11-03
Goose
An extensible open-source agent that can install, execute, edit and test with any LLM.
verified 2026-11-03
Hermes Agent
A self-hosted personal agent with long-term memory and self-written skills that you can reach from the terminal or from Telegram, Discord or Slack.
verified 2026-11-03
OpenHands
Self-hosting a team of coding agents, sandboxed in Docker, with scheduled or webhook-triggered automations.
verified 2026-11-03