cognis-code

One command → a local, uncensored AI coding suite wired into every IDE. No cloud, no keys, no limits.

#ai-coding #local-llm #uncensored #ollama #opencode #aider #continue #copilot-alternative

Download once, point VS Code, JetBrains, Cursor, Zed, Neovim, opencode, and Aider at the same local uncensored coder + reasoner. A private, unrestricted Copilot you fully own.

Usage — step by step

1. Install the CLI (console-script: cognis-code):

   pipx install "git+https://github.com/cognis-digital/cognis-code.git"
   cognis-code --version

2. Check your local setup and list the model roles it expects:

   cognis-code doctor
   cognis-code models

3. Pull a model (defaults to the coder role) and start the local OpenAI-compatible endpoint:

   cognis-code pull coder
   cognis-code serve            # serves on http://localhost:11434/v1

4. Wire an IDE/agent to that endpoint (use all to write every supported config). Preview first with --dry-run:

   cognis-code ide all --endpoint http://localhost:11434/v1 --dry-run
   cognis-code ide all

5. In CI, fail the job if the local setup is broken (doctor returns a non-zero exit on failure):

   cognis-code doctor || exit 1

Install (every way)

pip install "git+https://github.com/cognis-digital/cognis-code.git"   # or pipx / uv tool install
curl -fsSL https://raw.githubusercontent.com/cognis-digital/cognis-code/main/install.sh | sh
docker run --rm ghcr.io/cognis-digital/cognis-code --help

30-second setup

cognis-code pull coder        # pull an uncensored coder (Ollama) — or use uncensored-fleet
cognis-code serve             # local OpenAI-compatible endpoint at :11434/v1
cognis-code ide all           # write configs for EVERY IDE/agent at once
# now open VS Code/Cursor/Zed/opencode — they're already on your local model

Architecture

flowchart LR
  M[Uncensored coder + reasoner<br/>Ollama / llama.cpp / uncensored-fleet] --> E[OpenAI-compatible endpoint :11434/v1]
  E --> C[Continue · VS Code + JetBrains]
  E --> CU[Cursor]
  E --> Z[Zed]
  E --> N[Neovim · Avante]
  E --> O[opencode]
  E --> A[Aider]

Loading

What it wires up

IDE / Agent	How
VS Code · JetBrains	writes ~/.continue/config.json (Continue.dev)
Cursor	base-URL override snippet
Zed	~/.config/zed/settings.json assistant provider
Neovim	Avante/codecompanion config
opencode	~/.config/opencode/opencode.json local provider
Aider	~/.aider.conf.yml + local OpenAI base

Models (uncensored, swappable)

coder (Qwen2.5-Coder, abliterated) · coder-big (32B) · reasoner (DeepSeek-R1, abliterated) · commander (Josiefied-Qwen3-8B-abliterated). Powered by uncensored-fleet or plain Ollama.

Related

🤖 uncensored-fleet · 🧠 engram · 🔧 mcpify · 🗂️ the suite

⭐ Star it — own your coding AI, uncensored and local.

Responsible use

Local, unrestricted models are powerful. Use lawfully and ethically; you own what you generate.

Interoperability

cognis-code composes with the 300+ tool Cognis suite — JSON in/out and a shared OpenAI-compatible /v1 backbone. See INTEROP.md for the suite map, composition patterns, and reference stacks.

Integrations

Forward cognis-code's findings to STIX/MISP/Sigma/Splunk/Elastic/Slack/webhooks via cognis-connect. See INTEGRATIONS.md.

License

COCL v1.0 — see LICENSE.