Quecto Agent is set of extremely dense agents, with the most capable (agent_full.py) having:
- Stylized chat
- Full agent support for any provider
- Tool use
- MCP plugin support
- Reads AGENTS.md
- Skill discovery
All under 100 lines of Python
In a nutshell:
agent_fs.py: a singlefstool (list/read/write files under the working directory)agent_exec.py: a singleexectool (runs shell commands verbatim; Docker recommended)agent_exec_litellm.py: sameexectool, but uses LiteLLM so you can choose model/provideragent_full.py:exec(LiteLLM, Docker recommended) + AGENTS.md autoload + Skill discovery + MCP client loading
uv sync
export OPENAI_API_KEY="..."uv run agent_fs.pyModel is gpt-5.4 by default.
docker build -t quecto-agent -f docker/Dockerfile .These agents run arbitrary shell commands. Run them in Docker.
# OpenAI (gpt-5.4)
docker run -it --rm -e OPENAI_API_KEY quecto-agent agent_exec.py
# LiteLLM (OpenAI)
docker run -it --rm -e OPENAI_API_KEY quecto-agent agent_exec_litellm.py --model gpt-5.4
# LiteLLM (Anthropic)
docker run -it --rm -e ANTHROPIC_API_KEY quecto-agent agent_exec_litellm.py --model anthropic/claude-opus-4-7
# LiteLLM + extras (AGENTS.md + skills + MCP)
docker run -it --rm -e OPENAI_API_KEY quecto-agent agent_full.py --model gpt-5.4