Agent Mini
Local-first AI agent with shell, file, web and memory tools driven by small local models.
Tested on Linux
Use it
Linux x86_64
Install agent-mini via uv tool
uv tool install agent-mini
Check that it works
agent-mini --help 2>&1The output should contain “Commands:”.
What it does
- Run shell commands in your workspaceNot tested
- Read, write and edit filesNot tested
- Search the web and fetch pagesNot tested
- Save and recall facts in persistent memoryNot tested
How far pickfit got
Linux x86_64
- Installs
- Starts
- 0 of 4 features work
- Works offline
Good for
- Run a command in the workspace
- Read a file or line range
- Search the web via DuckDuckGo
- Fetch a public page as text
- Save and search persistent memory facts
How pickfit tested it
Each run starts from a clean machine and follows the steps above. How pickfit verifies
Linux x86_64
Its main function works
- Replayed from scratch
- Works offline
- Checked on 30 Oct 2026
- Version 0.5.0
Differences the test machine hit
- README: README providers say any OpenAI-compatible server works
In the test: use prebuilt llama-server tarball llama-b12015-bin-ubuntu-x64.tar.gz which extracts to ~/tools/llama/llama-b12015/llama-server, not build/bin/llama-server - README: no server setup documented
In the test: pip install llama-cpp-python fails here (no wheel build, needs toolchain); used prebuilt llama.cpp binary instead - README: config.example.json lists local model as local-model
In the test: local model id must be the server-reported id (/home/sbx/models/SmolLM2-135M-Instruct-Q8_0.gguf) for doctor Model check to pass