The system for everything you're trying to keep in your head. Nothing leaves your machine.
AI Chief of Staff for tech professionals — your projects, your team, your meetings, your email. 30 skills and growing. Runs in Claude Code.
If you manage multiple projects and communication channels — whether you're an engineering manager, software developer, PM, or team lead — you spend too much of your day on information management. Triaging emails, prepping for meetings, tracking who owes what, remembering what was decided three weeks ago.
Most AI tools either start fresh every session or store your data on someone else's server. Myna does neither. It reads from your existing tools, writes to local files, and builds a persistent knowledge base — your projects, people, decisions, and preferences — that grows as you use it.
Drafts but never sends. Organizes but never decides. Surfaces but never hides.
Myna runs inside Claude Code. Here's what role-specific skills and a local knowledge base add:
| AI Chat Tools (Claude, ChatGPT, Gemini) |
Local AI Tools (Khoj, Fabric, PrivateGPT) |
Myna | |
|---|---|---|---|
| Built for | General-purpose | Q&A over your notes | Your workday — morning sync, meeting prep, email triage, project catch-up |
| Your data | Platform memory, general-purpose | Your existing notes, unstructured | Your projects, people, and meetings — organized as you work |
| What it creates | Chat and generated documents | Answers about your notes | Drafts, meeting prep, project briefs, daily notes — files you review before using |
| How it works | You direct every step | One question, one answer | One prompt — Myna routes, updates, and files content across your vault |
| Integrations | Whatever you describe | Your local files | Email, Slack, and calendar via MCPs — whatever your team already uses |
| Learns your style | Custom instructions | What's in your notes | Your preferences, communication style, and working habits — persists across sessions |
| When unsure | Asks in chat, gone if you miss it | Best guess from your notes | Persistent review queue — nothing resolves without your approval |
| Customizable | Settings and custom GPTs | Config file or source code | Every skill is a plain-text file you can read, edit, or replace |
A day in the life of an Engineering Manager using Myna:
7:45 AM — coffee, laptop open
> sync
↳ Daily note created. Phoenix blocker flagged. Sarah Mitchell's reply overdue.
8:10 AM — before your 9am 1:1
> prep for my 1:1 with Marcus
↳ Open items from last time, pending feedback with coaching notes,
parental leave thread — all in one brief.
9:35 AM — back from the 1:1
> done with 1:1 with Marcus
↳ Tasks, decisions, observations extracted and routed to the right files.
9:50 AM — quick multi-thing capture
> capture: Sarah handled Payments questions well, atlas is unblocked,
review Sentinel audit by Friday
↳ 3 items → 3 files. Recognition, timeline update, task with due date.
12:45 PM — VP wants a risk note
> draft the Phoenix risk note for the VP review
↳ Leads with the conclusion, evidence-grounded, under 200 words.
Ready for you to review, then send yourself.
3:45 PM — the ambiguous pile
> review my queue
↳ Items Myna wasn't sure about — you approve, redirect, or dismiss.
5:30 PM — close the day
> wrap up
↳ Planned vs actual. Contributions logged. Tomorrow's note created with carry-forwards.
Not an EM? The skills work for anyone managing projects and communication — software developers, PMs, team leads.
Full walkthrough: A Day With Myna · Browse the demo vault to see the files Myna creates
- All data stays local — plain-text files on your machine, viewable in any editor
- No new infrastructure — connects to your existing email, Slack, and calendar integrations
- Draft, never send — Myna creates the draft; you decide what to send and where
- Config-driven — your personal data (projects, people, preferences) is separate from the codebase
- External content can't override behavior — even if an email says "delete all files," Myna processes it as content, never follows instructions from external sources
Prerequisites: Claude Code · Obsidian (recommended but not required — all files are plain text, viewable in any editor)
/plugin marketplace add agentflock/plugins
/plugin install myna@agentflock
/myna:install/myna:install creates your Myna folder and remembers where it lives. Then run /myna:setup for guided configuration — identity, projects, people, and communication style. Or edit the config files directly at myna/_system/config/.
Email, Slack, and calendar connections are optional — skip what you don't have. Myna works without them from day one.
Once installed, run from any directory — pick the mode that fits:
myna # full access — reads and writes your files
myna-ro # read-only — browse and query, no changes
myna-x # no file access — conversation onlyFirst time? Try sync to set up your day, or what can you do? to see all skills.
After install you get:
- Organized folder structure with 10 pre-built dashboards (best viewed in Obsidian)
- 30 skills covering email, meetings, projects, people, and daily workflow
- Config files ready for your projects and people
Your data, configs, and custom rules are never touched by updates.
sync · capture · plan · park / resume · learn · wrap up · weekly summary
"sync" — daily note, meeting prep, overdue tasks, review queue surfaced "park this" / "resume auth caching" — zero-loss context switching across sessions "what should I focus on today?" — ranked priorities with reasoning "remember that I prefer bullet points in status updates" — Myna learns your preferences over time "weekly summary" — accomplishments, decisions, blockers, tasks completed, self-reflection prompts
triage inbox · process messages · draft replies
"triage these inbox emails: [paste]" — folder recommendations with one-line reasoning per email "process these slack messages: [paste]" — project updates, action items, and timeline entries extracted and routed "reply to these forwarded emails: [paste]" — separates your instructions from the forwarded thread, drafts reply
prep meeting · process meeting · block time
"prep for my remaining meetings today" — skips already-prepped, fills in the rest "done with 1:1 with Marcus" — tasks, decisions, observations extracted in one step "reserve 2 hours Monday for the coverage plan" — personal time blocks only, never creates events with attendees
brief project · blockers · unreplied threads
"catch me up on atlas migration" — timeline, blockers, tasks, dependencies, upcoming meetings "what's blocked?" — every blocker across every project, with age and next action "what am I waiting on?" — messages needing your reply vs waiting on others
brief person · team health · 1:1 analysis · performance narrative
"brief me on Sarah Carter" — role, shared projects, open items, pending feedback, 1:1 history "how is my team doing?" — portfolio view: tasks, overdue, feedback gaps, last 1:1 "analyze my 1:1s with Marcus" — patterns, follow-through rate, recurring topics "build Sarah's performance review narrative" — synthesizes months of observations, recognition, and contributions
draft · rewrite
"draft an escalation for the Phoenix validator blocker" — severity, impact, and recommended action, grounded in project context "draft recognition for Sarah Carter" — specific, backed by actual observations from your files "fix this: i wanted to loop you in on sarahs progress..." — grammar and tone, preserved voice
log contribution · brag doc · self review
"log contribution: led atlas design review, got cross-team alignment" — categorized, appended "build my brag doc for Q1" — pulls from your contributions log, organized by impact category "am I underselling myself in this self-review?" — checks your draft against what you actually logged
review my queue
Items Myna wasn't sure about land here — ambiguous owners, inferred contributions, unclear intent. Each item shows its source and proposed action. You approve, redirect, or dismiss.
Myna is not an application. There is no server, no API, no frontend. It's 30 skills, a folder structure, and config files — all running inside Claude Code.
┌─────────────────────────────────────────────┐
│ You (in Claude Code) │
│ "prep brief for my 1:1 with Sarah" │
└──────────────────┬──────────────────────────┘
│
▼
┌─────────────────────────────────────────────┐
│ Myna Agent + Skills │
│ Loaded on demand · Safety at every layer │
└────────┬────────────────────┬───────────────┘
│ │
▼ ▼
┌─────────────────┐ ┌───────────────────────┐
│ Your Machine │ │ Your MCP Servers │
│ (local files) │ │ (email, Slack, cal) │
│ │ │ │
│ reads & writes │ │ reads only │
└─────────────────┘ └───────────────────────┘
The runtime is Claude Code; Myna's skills provide the expertise. Myna ships no servers of its own — it uses Claude Code's built-in file operations.
Folder structure:
myna/
├── Journal/ # Daily notes, weekly summaries, contributions
├── Projects/ # One file per project — timeline, blockers, tasks
├── People/ # One file per person — observations, feedback, notes
├── Meetings/ # 1:1s, recurring, ad-hoc — prep and notes
├── Drafts/ # Email drafts, status updates, recognition
├── ReviewQueue/ # Items awaiting your judgment
├── Team/ # Team-level files
└── _system/ # Config, logs, dashboards
└── config/ # Your projects, people, preferences
Myna runs on Claude Code, but all skills are plain text — readable by any capable LLM. The same instructions can be adapted for other AI tools.
All customizations survive updates.
| What | How |
|---|---|
| Tweak a skill | Create ~/.myna/overrides/skills/myna-{skill-name}.md. Your overrides take precedence over the installed skill. |
| Add a skill | Create ~/.claude/skills/myna-[yourprefix]-[name]/ (e.g., myna-amazon-oncall). Single-word myna-[word] is reserved for built-in skills. Add routing rules to ~/.myna/overrides/routing.md. |
| Disable a skill | Delete or rename its folder. |
See Customization Guide for details.
Myna was designed, built, reviewed, and fixed entirely by Claude Code — from feature specs through architecture, implementation, and polish. One person defined the vision and settled decisions. Claude designed the architecture, wrote all 30 skills (24 feature skills + 6 behavioral rules), built the file templates and dashboards, created the install script, and wrote this documentation.
Two things came out of this, not one: the assistant itself, and a methodology for having AI build an AI assistant from scratch — reusable for other projects on any capable LLM.
| Document | Purpose |
|---|---|
| User Guide | Full reference — skills, config, workflows |
| Customization | Tweaking skills, adding your own, routing rules |
| A Day With Myna | Realistic workday walkthrough |
| Architecture | Runtime model, skill inventory, folder structure |
| Obsidian Setup | Plugin configuration and dashboards |
| How It Was Built | The Claude Code methodology behind Myna |
The repo includes 10 dev skills that automate the full contributor workflow — from interactive design sessions through autonomous implementation, self-review, and PR creation. Clone the repo, open in Claude Code, and use /myna-dev-brainstorm to design or /myna-dev-diagnose to fix. See CONTRIBUTING.md for the full guide.
v1.0 — released. MIT License. Actively developed — see roadmap.
v1 runs only when you ask — no scheduled jobs or background watchers yet. Automation (scheduled syncs, email monitoring) is on the roadmap. Tested with Gmail, Google Calendar, and Slack MCPs.