chad: a local Claude-Code-style coding agent for your laptop

Claude can do anything, for anyone, anywhere. chad does one thing. 🗿 Coding under supervision.

chad is a single-user coding agent that runs entirely on an Apple Silicon Mac via MLX. One 27B model and no API key. (Not affiliated with Anthropic.)

Plenty of harnesses run local models now, and pi is a fantastic default for the same reason llama.cpp is: it works with everything. I was steering chad in the opposite direction. One model and one set of silicon, taken to the max. Swap out your CHAD_MODEL and it still runs, you just leave the drafter and the kernels behind.

Try it

uvx chad-code          # runs anywhere; the command is still `chad`
uvx chad-code prove    # offline smoke test: 4 tiny fix-it tasks, verified, timed 🗿

The first run asks, then downloads the model once (~8 GB) into the shared Hugging Face cache. While it downloads, cd into a project and think of a scoped first ask: "fix the failing test in tests/test_x.py" lands, "improve my codebase" flails.

chad targets 24 GB and nothing smaller. It runs below that and tells you it is doing so, but 8 GB of weights and drafter plus a ~4 GB prefill transient sit resident before a single token of context, so a 16 GB Mac gets a window too small to work in.

The PyPI package is chad-code. Bare chad is an unrelated squatted package.

Real session, unedited (the silent prefill is cut). A local model finds the cent that floor division loses, fixes it, and verifies itself, then runs your own pytest in your own shell. Recorded with --yolo so nothing pauses for a keypress; the default mode stops and asks before every edit and every command.

Same model, same Mac, stock engine

What do you gain over pointing a generic local-model tool at the same weights? Qwen3.8-27B at the UD-Q3_K_XL recipe (Unsloth's GGUF for llama.cpp, chad's MLX conversion of the same bit map; chad runs it with --model), the same M4 Pro (24 GB), one engine resident at a time, each measured with its own benchmark on a 512-token prompt and a 128-token generation.

A 200-token function body takes roughly 18 seconds at 10.9 tok/s and 3 at 63. You wait for the first one and you talk to the second.

The weights chad ships by default are the same model in Prism ML's ternary build (every projection Hadamard-rotated and stored at 2 bits): 7.2 GB resident instead of 12.3, the same 64 tok/s drafted and 21 serial on this Mac, and the 5 GB it gives back is context, roughly a 150k-token window where the 3-bit landed near 56k. The like-for-like table above stays on the 3-bit recipe because that is the quant llama.cpp has a GGUF of.

Ollama does not get its own row: it is llama.cpp underneath, measured without speculative decoding, and on the same GGUF (0.32.15, Modelfile FROM only) it measures 96 tok/s prefill and the same 10.9 decode.

¹ 63 is a ceiling: chad-bench's prompt is tiled code the drafter reads easily. Replayed against ten real mid-session contexts from ~/.chad/sessions (12-19k tokens, tool results in place, 384-token decodes) the same engine measures 31.7 tok/s median / 21.4 floor greedy against 14.8 serial, and 27.6 / 17.7 thinking against 13.9. That ~2× is what a session lives at.

² llama.cpp has run DFlash2 since build 10658. Here it accepts 96.5% of drafted tokens and still gains nothing: verifying 8 tokens costs ~6.4 serial steps on this GGUF and Mac (details).

Method, the longer runs and the caveats are in Throughput & performance; the rows are committed under benchmarks/stock/_runs/; reproduce them with uv run python benchmarks/stock/stock.py {llama,chad}.

Where the speed comes from

chad owns its inference loop instead of talking to a server, and the engine is fitted to the one checkpoint it ships:

• DFlash2 block speculation. A bundled 1.9B drafter (z-lab's DFlash2, ported to MLX and quantized) proposes a whole block of tokens from the main model's own hidden states. The main model verifies the block in one batched forward, and exact rejection sampling keeps every emitted token the model's own.
• A persistent prefix KV cache. The transcript is kept a strict token-prefix of the live cache, so a follow-up step prefills the ~16 tokens it appended instead of the 5,000 it already read: ~0.55 s per step instead of ~48 s. Any server with prompt caching gets the easy case; the work is holding it true across compaction, truncated turns and restarts. The system prefix is checkpointed to disk, so the second session anywhere starts warm (75.6 s → 5.5 s to the first tool call).
• Fused Metal kernels. Quantized-KV attention, a small-M matmul for speculative verify, and a compiled single-token layer step, chosen per machine at load time, no knobs.

What chad gives up

He has some of the same moves (tool use, plan mode, a real TUI) but he is a blunter instrument:

Five tools: bash, edit, write, write_todos, done. bash is the primary one because the model already knows rg and sed -n. The todo list is a checklist the model copies forward and ticks. Speed and the Claude Code muscle memory got the engineering budget; everything else is deliberately plain.

Interactive UX

uv run chad launches a full-screen terminal UI (built on prompt_toolkit):

• shift-tab cycles permission modes: normal (confirm each bash/write/edit) → auto-accept edits (edits land silently, terminal commands still ask) → yolo (nothing asks) → plan mode (read-only: investigate and propose a numbered plan) → back. A finished plan lands in ./plans/; ctrl-g (or /accept) clears the context and starts implementing it (details).
• Type-ahead message queue. Keep typing while the agent works; messages run in order.
• ctrl-c interrupts the running turn without killing the session. ↑prefilled / ↓generated token counts show an advancing % on an unavoidable full re-prefill, so it is never silent.
• @file / @dir mentions and !command shell passthrough. Pull a file into context inline, or run a shell command without invoking the model.
• Standing project instructions. A CLAUDE.md (or AGENTS.md) in the working directory is appended to the system prompt, and /init reads the project and writes one for you (details).
• Voice mode, all local. /speech, then ctrl-t to talk: Parakeet-on-MLX transcribes into the input box for you to review before Enter sends it, and replies are read aloud via macOS say. A word table teaches it your identifiers. Needs the speech extra (details).

uv run chad --help is the source of truth:

Two subcommands, each with its own --help: chad prove (the offline smoke test) and chad levers (print the result-channel lever registry as JSON, for A/B ablation).

A headless task (positional, or piped with no TTY) auto-approves mutating tools and runs greedy (temp 0). Every conversation is persisted under ~/.chad/sessions/, and every resume forks a new branch rather than overwriting.

The model

chad ships exactly one, downloaded once into the shared Hugging Face cache (~/.cache/huggingface, reused across every project). There is no picker and no size tier.

Qwen3.8-27B is dense (64 layers: 48 GatedDeltaNet + 16 full attention), so every parameter is on the critical path for every token and the quant is where both decode speed and context come from: the governor prices a gigabyte of weights at about 29k tokens of window. The ternary build is Prism ML's Bonsai pack of this checkpoint, repacked text-only with the base tokenizer and the drafter bundled (Apache-2.0; created using Bonsai by Prism ML). Its projections are stored in a rotated basis that an ordinary MLX loader would silently get wrong, so chad carries its own loader for it and fits the decode fast-path, the verify kernel and the drafter to it. What it costs: on code, teacher-forced perplexity is 4.49 against the 3-bit's 3.99 (+12%), while the private eval tiers tie at 56/56. The 3-bit stays one flag away.

--model <repo or local dir> runs different weights through the same engine and stays a first-class escape hatch. The drafter, the fused-attention coverage, the decode fastpath and the context governor are all fitted to the two checkpoints above, so other weights run slower; they do not break.

Installing & upgrading

The one-line quickstart (uvx chad-code) is up top. The other ways in:

uv tool install chad-code   # install for good, then it's just `chad`
uvx --from git+https://github.com/nathansutton/chad chad   # bleeding-edge main, no clone

Or from a clone (the dev path):

uv sync                      # install deps + the `chad` entrypoint (one time)
uv run chad                  # full-screen TUI
uv run chad "add a --json flag to main.py and update the tests"   # one-shot, headless
uv run chad -c               # resume this directory's last conversation

Two features are opt-in because they pull deps not every install wants: speech (voice mode: a mic library, no torch) and highlight (syntax colour in diffs and previews). An extra rides on the install spec, not on a separate command, so how you add it depends on how you installed chad:

uv tool install --force 'chad-code[speech]'   # add to an existing `uv tool` install
uvx --from 'chad-code[speech]' chad           # one-off run, nothing installed
uv sync --extra speech                        # from a clone

/speech in the TUI prints whichever of those matches your install, so you never have to work it out from here.

Upgrading depends on how you installed: uv tool upgrade chad-code, uvx --refresh chad-code, or git pull && uv sync for a clone. What changed lands in CHANGELOG.md. Model weights are versioned separately, so a code upgrade never re-downloads the model.

For development, uv sync once, then uv run pytest -q. The fast unit gate loads no model weights, runs in seconds, and is what CI runs. For throughput on your own machine, use uv run chad-bench (see Throughput & performance).

Extending chad

chad speaks the same two extension formats as Claude Code:

• Agent Skills. Drop a SKILL.md folder in ./.agents/skills/ or ./.claude/skills/ (or under ~/ for every project) and it becomes a slash command: /ship, /investigate the flaky test. Skills cost nothing until you run one, because chad puts no skill catalog in the system prompt. Precedence and the full list of roots are in the Configuration reference.
• MCP servers. Configure stdio or HTTP servers in ./.mcp.json to expose external tools (GitHub, Postgres, Linear, Slack, …) alongside chad's builtins, with static-token and OAuth auth.

Both are covered in full in the Configuration reference.

Documentation

• Design & internals covers why prefill is the bill, the persistent prefix cache, the trimmable/append-only trade, why the tool surface is five tools, and the ideas borrowed from other agents.
• Throughput & performance has the prefill, decode and warm-step numbers you can reproduce with chad-bench, the stock-engine comparison, and what the cross-session warm start is worth.
• Configuration reference documents project instructions, Agent Skills, MCP servers, plan mode, the slash commands, the context window, every environment variable, and the safety opt-outs.
• Troubleshooting maps symptoms to knobs for when a session rambles, loops, or slows.
• Contributing says what lands easily and what needs a conversation first.