Conifer

A quiet operating system for local AI.

---

Conifer runs local language models on your machine — model setup, storage, quantization, memory, and hardware-aware execution — so local AI feels fast, reliable, and hard to break. The engine runs in-process: no cloud roundtrip, no telemetry.

Download

Platform	Download	Notes
macOS · universal	Conifer_universal.dmg	Apple silicon + Intel · macOS 11+ · Metal
Windows · x86-64	Conifer_x64-setup.exe	Windows 10+ · Vulkan/CPU
Linux · x86-64	Conifer_amd64.AppImage	.deb also attached

Once installed, Conifer keeps itself current: it checks for new releases on launch, downloads them in the background, and offers a one-tap Restart to apply — never interrupting a running model or agent. You never re-download by hand.

Models are downloaded inside the app the first time you use them. The installer ships the engine, not the weights, so it stays small.

What you get

• Local inference, in-process. The model runs on your CPU/GPU. Nothing leaves the machine.
• Hardware-aware execution. Metal on Apple silicon; Vulkan and CPU on Windows and Linux, with NVIDIA CUDA arriving as an automatic update.
• A real model manager. Browse, download, quantize, and store models without hand-editing config.
• An OpenAI-compatible server you can point your existing tools at — see building on top of Conifer.

Building on top of Conifer

Conifer exposes a local, OpenAI-compatible HTTP API (and an Ollama-native one), so anything that speaks those protocols — coding agents, editor extensions, your own scripts — can use Conifer as a drop-in local backend. No account, no key, no cloud.

# Point any OpenAI client at the local server:
export OPENAI_BASE_URL=http://127.0.0.1:8080/v1
export OPENAI_API_KEY=local        # ignored, but most clients require one

from openai import OpenAI
client = OpenAI(base_url="http://127.0.0.1:8080/v1", api_key="local")
resp = client.chat.completions.create(
    model="conifer",
    messages=[{"role": "user", "content": "Explain bandwidth-bound decoding."}],
)
print(resp.choices[0].message.content)

Full API reference: docs/api/ — the OpenAI endpoints, the Ollama-native /api/* surface, fill-in-the-middle for code completion, and the conifer serve CLI.

A note on what's open

The download is a compiled application. The high-performance inference engine and the agent runtime are Conifer's core technology and are not open source. What is open — and what this repository hosts — is everything you need to use and build on Conifer: the downloads, the release history, and the full API. That's the deal: a closed, fast core with an open, documented surface. See SECURITY.md and CONTRIBUTING.md.

Links

• Website & docs: conifer.build
• Releases & changelog: RELEASES.md · Releases
• Report an issue: open an issue

License

Repository contents (docs, API reference) are MIT. The Conifer application binaries are proprietary under the Conifer End-User License — install and use them freely; don't redistribute, reverse-engineer, or repackage them.

_{Built by ConiferKit. Conifer is a quiet operating system for local AI.}