Native for Apple Silicon · macOS 14+

Write code with AI, on your own machine.

Aidesmith is a macOS app that lets you build software with AI while keeping most tokens on your own machine. Tell it what you want by voice or text. It plans the work, runs it locally, checks the result, and commits each finished piece to git.

Request early access See how it works

40–60% fewer cloud tokens. Local handles the heavy lifting.

Aidesmith workbench: Architect chat, task queue, execution log, and live code view

Aidesmith menu bar tooltip showing remote quotas and local usage

The core idea

Cloud models are great at thinking. Let them do just that.

Writing, retrying, and checking happens on your Mac. No meter running.

Before

100%

All tokens in the cloud

With Aidesmith

40–60%

Cloud + local split

Architect / Executor

A proven pattern, taken further.

A cloud model understands the codebase and breaks the work into small tasks with clear acceptance criteria. A local model carries each one out in an agentic edit loop. The token-hungry part runs on hardware you already own.

Architect

Cloud

Cloud · Anthropic, OpenAI, Google

Talks with you, reads the codebase, breaks work into small tasks with acceptance criteria.

Planning
Code understanding
Judgment calls

Executor

Local

Local · Ollama or compatible providers

Carries each task out in an agentic edit loop on your machine. File reads, retries, corrections.

Edit loop
Retries
Token-heavy work

Ralph-style loops

Done when checks pass, not when the model says so.

Every task is validated mechanically with greps, lint, and the project's own tests. Failures feed straight back into the next attempt. Retries are bounded and informed by the actual failure.

# task: add refresh token rotation

executor → wrote src/auth/token.service.ts

validate → npm test -- token

✗ rotateRefreshToken: payload.jti undefined

executor → retry with failure context

validate → 4 passed, 0 failed

lint → 0 problems

git commit -m 'add refresh token rotation'

Project awareness

Plans built against the real codebase, not assumptions.

When you open a project, Aidesmith studies it on device — no source leaves your machine for indexing.

Semantic index

Local embeddings scan and index the codebase for meaning-based search.

Structural analysis

Tree-sitter extracts types, functions and signatures across any language.

File synopses

Apple Intelligence writes a one-line summary for every file, on device.

Hard facts

Language, framework version, test commands, and your coding conventions.

Features

Built for shipping, not demos.

Three-pane workbench

Chat with the Architect, watch the task queue, inspect every execution. Flip on the code view for a live diff of every change.

Bring your own models

Anthropic, OpenAI-compatible, Google, or local providers like Ollama. Pick per role.

Quota & usage tracking

Per-task cloud/local token split with provider quota gauges in the menu bar.

Auto-pause & auto-resume

Hit your quota limit? Work pauses safely and picks back up automatically when quota returns.

One commit per task

Every task that passes validation gets its own clean git commit.

Coming Soon

Auto or supervised

Per-pane mode with force-run or pause for individual tasks.

Menu bar quota tooltip

Live remote quotas and local model usage, one click from anywhere on macOS.

FluidAudio voice chat

Talk to Aidesmith naturally with Parakeet ASR and SenseVoice. Streaming, low latency, entirely on device.

Specs

The hard facts.

Platform: Native SwiftUI · Apple Silicon · macOS 14+
Local models: Ollama and other local providers
Cloud models: Anthropic, OpenAI-compatible, Google · your own keys
Speech: FluidAudio voice chat with Parakeet ASR and SenseVoice · on device
File synopses: On-device via Apple Intelligence
Semantic index: Local embeddings + tree-sitter symbol extraction
Secrets: API keys in the macOS Keychain
Version control: Git, one commit per validated task

Stop paying cloud rates for every token.

Aidesmith is in private beta. Request access and we'll be in touch.

Request early access →