A starter workspace for running a team or company with Claude Code.
You clone this repo, open it in Claude Code, type start here, and the built-in onboarding agent walks you through a 10-minute setup. By the end you have a workspace that captures your work, tracks your goals, and saves itself to git on demand.
This is the entry point for the Ace system — the no-frills version. Bring your code or your business operations into this repo, and Claude meets you where you work.
| CLAUDE.md context | The brain. Claude reads it every session. Defines structure, conventions, and how to behave in this repo. |
| Hooks for logs | Python hooks that nudge every chat to capture work as it happens — temp logs that auto-consolidate into monthly session logs. |
| Tasks inbox | A single open list (tasks-inbox.md) for anything that doesn't yet have a home. Sort it later when goals get set up. |
| Agents that manage and consolidate context | session-logger (rolls temp logs into monthly logs), status-reporter (dashboard across projects), workspace-sync (git through plain English). |
| OKR onboarding agent | goals-onboarding walks you through annual objectives → quarterly KRs → initiatives → tasks, all stored as YAML. |
| Sync agent to save team work | workspace-sync commits and pushes through natural language — "save my work", "what changed since Monday", "undo the last change". |
| VS Code setup agent | vscode-setup configures the editor, the recommended extensions, the workbench layout, and explains the permission layers (Claude in VS Code vs Claude in terminal). |
If you don't have it yet, grab it at https://code.visualstudio.com — Cursor, Antigravity, and VSCodium also work.
On the VS Code welcome screen, choose Clone Git Repository, paste this URL, and pick a folder:
https://github.com/41fred/ace-level1.git
VS Code will offer to install the recommended extensions — say yes. The Claude Code extension is the important one.
Prefer the terminal? Same result:
git clone https://github.com/41fred/ace-level1.git my-workspace
cd my-workspace && code level1.code-workspaceIn the Claude Code chat panel, type:
intro
That runs the system-overview agent — a five-minute walk through how Ace works (the problem, the three pillars, the business model, the core agent roles, the result). It's the recap of Alfred Naayem's Ace presentation built straight into the workspace, and it hands off to setup when you're ready.
Skip it if you want to jump straight in.
In the Claude Code chat panel, type:
start here
That wakes the built-in onboarding agent, which walks you through:
- Identity (your GitHub username, name, git config)
- Which agents to keep active
- Permission mode (run free, ask each time, or allowlist specific tools)
- MCP integrations (Google Workspace for calendar — optional)
- Goals (hand-off to
goals-onboardingif you want to set up OKRs) - Connecting other code repos if you have them
By the end, your README.md and CLAUDE.md reflect what you actually have configured.
You're done with setup. Open a file, ask Claude a question, build something.
A few phrases that work out of the box:
| Say | What happens |
|---|---|
save my work |
Commits and pushes through workspace-sync |
what's the status |
status-reporter shows active priorities and blockers |
consolidate my logs |
session-logger rolls temp logs into the monthly session log |
weekly review |
goals-weekly-review collects KR actuals and drafts a CEO email |
what should I work on |
pm triages your tasks into ready / blocked / needs-decision |
find patterns |
skill-mining surfaces wins, friction, and SOP candidates from history |
check my calendar |
calendar-audit reviews scheduled events and logs outcomes |
set up VS Code |
vscode-setup walks you through editor configuration |
Ace Level 1 is the smallest workspace that proves the loop works:
- Your data stays in your repo. Markdown, YAML, and a few Python hooks. No external SaaS, no vendor lock-in. Move it anywhere.
- Claude gets smarter with use. Every session writes notes that Claude reads on the next session. The repo becomes a cumulative memory.
- The things that work get repeated. Patterns you discover (a good prompt, a recurring weekly task, a checklist that catches bugs) live in
.claude/agents/and run the same way every time. - Your processes are absorbed. SOPs become files Claude reads. New team members run
start hereand inherit your conventions.
It's deliberately small. When you outgrow it — multi-repo portfolios, custom hooks, MCP-driven analytics, daemon-scheduled tasks — there's a Level 2.
ace-level1/
├── README.md ← you are here
├── CLAUDE.md ← workspace brain — read on every session
├── tasks-inbox.md ← open list of unsorted work
├── goals-schema.md ← reference spec for OKR YAML
├── level1.code-workspace ← VS Code workspace file
├── .claude/
│ ├── agents/ ← installed agents (manifest.json lists them)
│ ├── hooks/ ← session-init.py, prompt-check.py
│ ├── rules/ ← always-active behavioral rules
│ └── settings.json ← permission mode + hook wiring
├── .vscode/
│ ├── extensions.json ← recommended VS Code extensions
│ └── settings.json ← workspace settings (markdown preview, etc.)
├── logs/
│ ├── temp/ ← real-time temp logs (auto-consolidated)
│ └── archive/ ← processed temp logs (excluded from search)
└── .gitignore
| Agent | Status | One-liner |
|---|---|---|
system-overview |
always available | Tour of how Ace works — recap of Alfred Naayem's Ace presentation |
start-here |
always available | First-run setup walk-through |
vscode-setup |
always available | Configure VS Code: extensions, layout, permission layers |
session-logger |
default | Consolidate temp logs into session logs |
workspace-sync |
default | Git operations through natural language |
status-reporter |
default | Project dashboard |
calendar-audit |
optional | Review calendar events (needs Google Workspace MCP) |
goals-onboarding |
optional | Set up OKR YAML structure |
goals-weekly-review |
optional | Weekly KR actuals + CEO narrative email |
pm |
optional | Triage tasks into ready / blocked / needs-decision |
skill-mining |
optional | Surface patterns — wins to repeat, friction to fix, SOP candidates |
Read .claude/agents/manifest.json for the full inventory.
Claude Code has two permission surfaces — the VS Code extension and the terminal CLI. Both read .claude/settings.json from this repo.
This starter ships with:
{ "defaultMode": "bypassPermissions" }Meaning Claude runs without permission prompts inside this repo. You review changes via git diff before pushing.
If you want Claude to ask before each action instead, run the vscode-setup agent and pick "ask each time" — or edit .claude/settings.json directly.
vscode-setup is also where to go if you want to allowlist specific tools (e.g., let Claude run git status and npm test without asking, but still confirm everything else).
Run goals-onboarding when you're ready to set up structured goals. It walks you through:
- Annual objectives (3–5 outcomes for the year)
- Key results for each (measurable targets)
- Quarterly OKRs (optional — if a team has its own breakdown)
- Initiatives and tasks (the projects driving the objectives)
- Org chart (who owns what)
Output is a goals/ folder of YAML files following goals-schema.md. Once set up, run weekly review (the goals-weekly-review agent) to update actuals and generate a CEO-ready email.
MIT. Use it, fork it, ship it.
When you outgrow L1, one piece of infrastructure is open-sourced and ready today:
github.com/41fred/claude-code-slack
A Slack ↔ Claude Code bridge. Send a message in Slack → Claude Code runs in your repo → result posts back in Slack. Includes scheduled task support so agents like skill-mining and goals-weekly-review can run on cron without you opening the editor.
Pairs with L1 — your repo doesn't have to change. The daemon executes against it.
- Level 2 — dashboards + Telegram/Slack-native experience
- Level 3 — database, multi-user interface, role-based access
Might release in future versions — stick around.
Ace is built by Alcanah Partners, founded by Alfred Naayem.
Questions, feedback, or stuck on something? Email alfred@alcanah.co — Alfred reads everything.