Registry discovery (drone, spawn, seedgo) now globs for *_REGISTRY.json with CWD-first walk-up — enabling project isolation without hardcoded filenames. Drone stderr crash fixed (Rich Console). Pyright OOM fixed via pyrightconfig excludes. AIPASS.md simplified for universal use. Global prompt updated with flow commands. DPLAN-003 added for AIPass-as-OS design vision. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
AIPass
Building in public. Active development — not a finished product. Expect breaking changes.
Orchestration framework for autonomous AI agent ecosystems. Command routing, symbolic addressing, standards enforcement, workflow management, and inter-agent messaging. Agents use @branch names that resolve at runtime instead of hard-coded paths. Each agent has persistent identity and memory via .trinity/ files (passport, session history, observations).
Quick Start
1. Install
git clone https://github.com/AIOSAI/AIPass.git
cd AIPass
./setup.sh
source .venv/bin/activate
setup.sh handles everything: creates the venv, installs the package, generates the branch registry (15 branches), bootstraps identity files (.trinity/, .seedgo/, .ai_mail.local/) for every branch, and installs Claude Code hooks. Idempotent — safe to re-run.
Verify:
drone systems # Should show 15 branches
drone @seedgo verify # Should show 5/5 checks passed
2. Start a Session
AIPass is designed to be operated by AI agents via Claude Code. The orchestration hub is devpulse — start there.
cd src/aipass/devpulse
claude --permission-mode bypassPermissions
Say hi to trigger the startup protocol. The agent reads its identity and memory files, checks git status, verifies system health, and picks up where it left off. In a returning session, use /resume instead.
3. What You Get
After setup, every branch has:
.trinity/passport.json # Identity (name, role, citizen class)
.trinity/local.json # Session history (empty, ready to populate)
.trinity/observations.json # Collaboration patterns (empty)
.seedgo/bypass.json # Standards bypass config
.ai_mail.local/inbox.json # Mailbox
The agent fills in its own memories as it works — session logs, learnings, observations. These grow over time and persist across sessions.
4. Startup Protocol
These greetings trigger the full startup sequence: hi, hello, yo, hey, sup, good morning, good evening, what's up. Everything else is treated as a direct task.
On startup the agent:
- Reads
.trinity/passport.json(identity),local.json(session history),observations.json(patterns) - Runs
git status,drone systems,drone @seedgo verify - Checks active tasks and recent session context
- Picks up work or waits for instructions
Core Concepts
Drone — Command Router
Everything goes through drone. It resolves @branch names to paths and routes commands.
drone @seedgo audit aipass # Run standards audit
drone @seedgo verify # Check seedgo health
drone @spawn --help # Show help for a branch
drone systems # List all registered modules
In Python:
from aipass.drone.apps.modules.resolver import resolve_branch
from aipass.drone.apps.modules.registry import load_registry
path = resolve_branch("@drone") # Resolve @name to path
registry = load_registry() # Load full branch registry
Seedgo — Standards Enforcement
Standards-based code auditor. Checks are organized into packs — the built-in aipass pack has 21 standards covering imports, architecture, naming, error handling, logging, and more.
drone @seedgo verify # Health check
drone @seedgo list # Show installed standard packs
drone @seedgo audit aipass # Run the aipass standards pack
seedgo checklist aipass file # Check a single file
Symbolic Addressing
All modules register in AIPASS_REGISTRY.json. Drone resolves @name to paths at runtime — no hard-coded paths between modules.
Architecture
All modules follow a 3-layer pattern:
src/aipass/<module>/
├── apps/
│ ├── <module>.py # Entry point (what drone routes to)
│ ├── modules/ # Business logic
│ └── handlers/ # Implementation details
15 branches — all under active development. Ported from a private predecessor project; not all modules are fully tested or wired up yet.
| Branch | Purpose |
|---|---|
drone |
Command routing, @branch resolution |
seedgo |
Standards enforcement, 21-standard audit pack |
prax |
Logging and monitoring (the only logger) |
cli |
Display formatting (Rich) |
flow |
Workflow and plan management (FPLAN lifecycle) |
ai_mail |
Inter-agent messaging, dispatch, wake |
spawn |
Branch lifecycle — create, update, delete |
trigger |
Event-driven automation, circuit breaker |
api |
LLM access via OpenRouter, key management |
backup |
Multi-mode backup (snapshot, versioned, Google Drive) |
daemon |
Background scheduler, cron, Telegram notifications |
memory |
Vector memory bank (ChromaDB, sentence-transformers) |
commons |
Social network for branches — posts, rooms, artifacts |
skills |
Capability framework — discoverable, executable skill units |
devpulse |
Orchestration hub (no code — coordination only) |
Docker
docker build -t aipass-test .
docker run -d -p 8080:8080 --name aipass-vscode aipass-test
Opens VS Code in browser at http://localhost:8080 with AIPass pre-installed.
Requirements
- Python 3.10+
- No external API keys required
- Dependencies:
rich,watchdog
License
MIT