Steer your projects with AI
A frame that splits work into project, epic, story and task and moves it forward by phases, with the agent as a teammate rather than a black box.
145 MCP verbs
ArboRithm
Steering AI agents, giving them memory and tools.
ArboRithm builds tools for AI-assisted work: a project pipe and a persistent memory.
The craft
Organising your work with artificial intelligence: four capabilities, from project steering to the interface.
Give agents a memory
An agent forgets everything between sessions. A vault of readable Markdown files gives it back the context, the decisions and their reasons.
some forty MCP tools
Let agents see and act
A local tool that lets any AI assistant see the screen and act on it. What it is allowed to do is decided by the tool, never the AI.
5 interaction levels
Design interfaces
The Monolith design system, dual-mode and accessible, is a craft that can also be handed over: the same tokens in night and day.
21 component folders
The projects
Three tools for working with AI agents, each with its own identity.
Effortless
Steer your projects with AI, effortlessly
Effortless is a framework for AI-assisted project steering. It organises work as Project, Epic, Story and Task, and moves it forward through five phases: Observe, Position, Articulate, Launch, Execute. The desktop app, local-first, exposes the engine through 145 MCP verbs. Free for personal or professional use: your projects stay yours, and no data is collected.
- A five-phase pipe, from idea to shipped code
- A fractal model: Project, Epic, Story, Task
- A team loop: analyst, developer, QA
- Agent-agnostic: Claude Code, Codex, Antigravity
- Per-module delivery: own version, repository and tag
- macOS (Apple Silicon) and Windows (x64, ARM64)
Kesshō
The memory that carries over between sessions.
An AI agent forgets everything between sessions. Kesshō gives its memory back: it writes and reads a vault of Markdown files, readable in Obsidian, and exposes it to agents through some forty MCP tools. Six weeks later, you find not only what was done, but what was decided and why.
- A readable Markdown vault, never an opaque database
- Dated archives and a current context per project
- Works with common agents through MCP
- A single native binary (Tauri, Rust)
%20scale(0.68)'%3e%3cclipPath%20id='daimon-clip-a'%3e%3cpath%20d='M344.17%20335.00%20C331.15%20366.70%20286.23%20397.15%20253.00%20406.42%20C219.77%20415.70%20174.43%20408.66%20144.78%20390.65%20C115.13%20372.64%2077.59%20330.51%2075.09%20298.37%20C72.60%20266.22%20103.71%20221.53%20129.82%20197.79%20C155.93%20174.05%20198.20%20152.84%20231.75%20155.91%20C265.30%20158.97%20312.37%20186.34%20331.11%20216.18%20C349.85%20246.03%20357.19%20303.29%20344.17%20335.00Z'/%3e%3c/clipPath%3e%3cpath%20d='M344.17%20335.00%20C331.15%20366.70%20286.23%20397.15%20253.00%20406.42%20C219.77%20415.70%20174.43%20408.66%20144.78%20390.65%20C115.13%20372.64%2077.59%20330.51%2075.09%20298.37%20C72.60%20266.22%20103.71%20221.53%20129.82%20197.79%20C155.93%20174.05%20198.20%20152.84%20231.75%20155.91%20C265.30%20158.97%20312.37%20186.34%20331.11%20216.18%20C349.85%20246.03%20357.19%20303.29%20344.17%20335.00Z'%20fill='%23B66CFF'/%3e%3cg%20clip-path='url(%23daimon-clip-a)'%3e%3cellipse%20cx='175.0'%20cy='231.4'%20rx='123.5'%20ry='110.5'%20fill='%23ffffff'%20opacity='0.2'/%3e%3cellipse%20cx='242.6'%20cy='361.4'%20rx='136.5'%20ry='101.4'%20fill='%2302101a'%20opacity='0.26'/%3e%3c/g%3e%3cclipPath%20id='daimon-clip-b'%3e%3cpath%20d='M409.98%20204.25%20C398.29%20213.06%20375.46%20214.37%20359.64%20210.67%20C343.82%20206.97%20321.17%20195.62%20315.06%20182.07%20C308.95%20168.52%20315.03%20143.33%20322.98%20129.36%20C330.94%20115.40%20347.35%20101.95%20362.78%2098.27%20C378.20%2094.58%20404.35%2097.34%20415.52%20107.26%20C426.69%20117.19%20430.71%20141.64%20429.79%20157.81%20C428.87%20173.97%20421.67%20195.44%20409.98%20204.25Z'/%3e%3c/clipPath%3e%3cpath%20d='M409.98%20204.25%20C398.29%20213.06%20375.46%20214.37%20359.64%20210.67%20C343.82%20206.97%20321.17%20195.62%20315.06%20182.07%20C308.95%20168.52%20315.03%20143.33%20322.98%20129.36%20C330.94%20115.40%20347.35%20101.95%20362.78%2098.27%20C378.20%2094.58%20404.35%2097.34%20415.52%20107.26%20C426.69%20117.19%20430.71%20141.64%20429.79%20157.81%20C428.87%20173.97%20421.67%20195.44%20409.98%20204.25Z'%20fill='%23E8B23A'/%3e%3cg%20clip-path='url(%23daimon-clip-b)'%3e%3cellipse%20cx='356.0'%20cy='130.8'%20rx='57.0'%20ry='51.0'%20fill='%23ffffff'%20opacity='0.2'/%3e%3cellipse%20cx='387.2'%20cy='190.8'%20rx='63.0'%20ry='46.8'%20fill='%2302101a'%20opacity='0.26'/%3e%3c/g%3e%3c/g%3e%3c/svg%3e)
Daimon
See, act, show: any AI on your desktop, safely.
Daimon lets any AI see your screen, act on it and show you what it is doing. One local daemon, on macOS and Windows, opens these three capabilities to it through the MCP protocol. The hands stay under a ceiling that Daimon enforces itself, never the AI, and that starts at L0, hands off.
- Three organs in one local daemon
- A hands ceiling from L0 to L4, held by Daimon and not by the AI
- Secrets blanked before anything is served, stoppable at any time
- Any MCP client, no lock-in, on macOS and Windows