v0.1.0 released — open source on GitHub

Persistent Memory &
Code Understanding
for Claude Code

An open-source orchestrator that gives Claude Code a persistent knowledge graph and code graph across your projects. The VCT Launcher is the hub that installs and connects everything. The Orchestrator wires up agents, skills, and hooks. Runs locally on your machine — open source, available now under AGPL-3.0.

View on GitHub →
Open source Runs locally Persistent memory Works with Claude Code More tools coming
Terminal - Orchestrator Setup
YouSet up my orchestrator
SystemStarting setup wizard. Detecting your system...
System Phase 1 - Linux detected, Python 3.12, Docker running
System Phase 2 - Local models: pulled
System Phase 3 - Vector DB: running on :8081
System Phase 4 - MCP servers configured
System Phase 5 - Knowledge graph ready
19
Agents (free tier)
28
Skills installed
20
Automation hooks
KG
Knowledge graph ready
19 agents, 28 skills, 20 hooks — numbers from the repo, not marketing
Knowledge graph across projects
Numbers come from the repo, not from marketing
Semantic code graph
Persistent memory across sessions
20 hooks, all open source
Local-first by default
MCP-native for Claude Code
Actually open-source (AGPL-3.0)
Embeddings stay on your machine
Built in Italy, shipped on GitHub
Find a discrepancy? Open an issue
19 agents, 28 skills, 20 hooks — numbers from the repo, not marketing
Knowledge graph across projects
Numbers come from the repo, not from marketing
Semantic code graph
Persistent memory across sessions
20 hooks, all open source
Local-first by default
MCP-native for Claude Code
Actually open-source (AGPL-3.0)
Embeddings stay on your machine
Built in Italy, shipped on GitHub
Find a discrepancy? Open an issue
The Ecosystem

Tools that work better together

Each tool is standalone and useful on its own - but they connect. The VCT Launcher is the hub: one place to install, activate, and launch everything. Tools share a common backend - a knowledge graph, a code graph, and a set of MCP servers - so the Orchestrator remembers context across sessions and connected tools reuse the same memory. Everything runs locally. Fully open source under AGPL-3.0.

View source on GitHub →

VCT Launcher as hubKnowledge graph memoryTools share contextFully local - no cloudOpen source (AGPL)More tools coming
Features

What's Actually
Inside the Box

What it does day to day.

Memory that persists across projects

Most AI coding assistants reset between sessions. The Orchestrator doesn't. Every decision and pattern you capture lands in a local knowledge graph — embedded on your machine, no cloud step. The graph is shared across projects, so the JWT pattern you documented last month is searchable from the codebase you opened this morning. When the session resets, a snapshot preserves your active state and restores it automatically.

Code the AI can query, not just read

Source files are parsed, not just embedded. Tree-sitter extracts an AST across 10+ languages and builds typed collections — functions, classes, modules, APIs, and cross-service calls (with protocol and confidence). Optional deep analysis (opt-in, requires a separate JVM component) adds control-flow and data-flow depth. Queryable via MCP: ask "what calls validate_token?" and you get a structured answer in milliseconds. Re-indexed automatically on every code edit, so it never drifts out of sync.

Automation as enforcement, not advisory

Shell commands are checked before execution, credentials are scanned on every write, and network fetches are guarded against internal IP access. Every subprocess starts with secrets stripped from its environment. The model cannot bypass any of this — enforcement is in the runner, not in the prompt. Before an Edit, the system pulls relevant patterns and decisions from the knowledge graph and injects them into context — just-in-time recall, not eager loading.

Local-first, your data, your machine

Weaviate runs in a container on localhost. Two embedding models — one for text, one for code — selected automatically based on your hardware. GPU gives you the larger code model; CPU falls back gracefully. Nothing crosses the wire to a third party unless you explicitly opt in. Secrets stay in a local vault, scoped per project: one command runs any subprocess with the right credentials injected — nothing echoed to the terminal.

Semantic Memory
Knowledge Graph Search
Ask "how did we handle auth in the last project?" and get the exact file, the pattern, and the decision rationale, pulled from any project in your graph. Always included in the free tier.
Parallel Execution
Agents that work in parallel
19 agents in the free tier, 10 more with MAO. Each runs in its own Claude Code context. While the coder implements a feature, the tester writes tests against it in parallel, not in sequence.
planner
Task decomposition
coder
Implementing auth module
tester
Writing test suite
doc-maintainer
Updating API docs
code-reviewer
Waiting for coder...
Automation
20 Automation Hooks
20 hooks handle credential scanning, knowledge graph auto-sync on edits, code graph updates, pre-compact context saves, and more. Events fire automatically as you work.
15:04:02[SSRF Guard]KG auto-sync: knowledge/concepts/auth.md
15:04:07[SSRF Guard]BLOCKED: 192.168.1.1 (private IP)
15:04:11[BAP]Code graph updated: src/api/routes.py
15:04:18[BAP]Pre-compact snapshot saved
15:04:22[Cred Scan]BLOCKED: AWS key detected in .env
15:04:30[Shell Guard]Command safe: git status
Instant Expertise
28 Slash-Command Skills
Type /architect and get a structured design review. /security-reviewer runs a full OWASP sweep. /tdd writes failing tests first, then the implementation. One command per workflow, no prompt engineering needed.
/architect/tdd/api-designer/security-reviewer/debug-expert/ai-rag-advisor/performance-optimizer/doc-template/ai-model-selector/explore-codebase+19 more...
Code Understanding
Code Graph the AI Can Query
Parses 10+ languages into typed collections — functions, classes, APIs, cross-service calls. Optional deeper analysis adds control-flow and data-flow. Queryable via MCP on every edit.
Just-in-Time Context
Hooks Inject Knowledge Before You Edit
When the agent is about to edit a file, the system pulls the most relevant patterns, decisions, and similar code from the knowledge graph and injects them into context. The system picks how much detail to inject — enough to help, not so much that it crowds out your code.
1
Edit src/auth/login.py
2
PreToolUse fires - hybrid_search()
3
3 KG nodes found, RL-scored
4
Injected: JWT pattern + decision + similar fn
Sandbox & Safety
Hooks Catch Mistakes Before They Happen
Shell injection, leaked credentials, and internal IP access — all blocked before the call reaches the tool. Env scrubbing on every subprocess. Hooks block, not just warn.
Bash(rm -rf /) → BLOCKED
Write(.env) AWS_SECRET_ACCESS_KEY → BLOCKED
WebFetch(192.168.1.1) → SSRF guard → BLOCKED
Bash(git status) → OK
Edit(src/auth.py) cred scan → OK
Per-Project Secret Isolation
Your Tokens Never Leak Into Code
Secrets stay in a local vault, scoped per project. One command runs any subprocess with the right credentials injected — nothing echoed to the terminal. Git credential helpers and MCP wrappers all read from the same vault.
$ vct exec --project orchestrator \
    --secret openai_key=OPENAI_API_KEY \
    -- python main.py
✓ subprocess started, env scrubbed
~/.vct-secrets/
├─ shared/github_pat
└─ projects/orchestrator/
└─ openai_key
MCP-Native, Pro Reranker
Four MCPs by Default + Pro RL Reranker
Four MCP servers out of the box: knowledge graph, local inference, web/code search, and code embeddings. Any third-party MCP works alongside. Pro adds a reranker that learns your codebase and adapts in hours, picking detail tiers automatically.
weaviate-kg
ollama
search
code-embedding
RL reranker accuracy
0h → 24h
Beyond One Repo
Knowledge that Travels with You
Patterns from one project stay searchable across all your projects — no manual cross-referencing. Decisions carry forward automatically. Most tools are project-local; the orchestrator is cross-project by default.
Acme API
Acme Mobile
Customer Portal
Shared KG
#auth-patterns#redis-cache#rate-limit
Productivity
Desktop Apps in the Ecosystem
Transcrypt transcribes audio locally. LineArtStudio converts photos to SVG vectors. ConvertiFacile handles PDF, audio, and image formats. Each app runs standalone. Install via VCT Launcher to share the orchestrator backend.
VCT Launcher
🎙
Transcrypt
LineArtStudio
📊
SlideMaker
ConvertiFacile
VCT Launcher
Orchestrator
Stack
CCClaude Code
AnAnthropic
VSVS Code
PyPython
DkDocker
OlOllama
WvWeaviate
GHGitHub
PdPodman
uvuv / pip
CCClaude Code
AnAnthropic
VSVS Code
PyPython
DkDocker
OlOllama
WvWeaviate
GHGitHub
PdPodman
uvuv / pip
How It Works

Up and Running
in Under 30 Minutes

No CLI setup scripts. No config files to edit. The wizard handles install, dependencies, and the first sync.

/1
Install & Open
Install the VibeCoded Orchestrator via the VCT Launcher and open your project in VS Code. The setup wizard loads automatically on first open.
/2
Wizard Configures
The installer detects your system, pulls local AI models, starts the knowledge graph, and copies agents + skills into your project.
/3
Use Everywhere
Agents and skills work inside Claude Code. The knowledge graph remembers across sessions. Tools you add via the launcher share the same backend.
AGPL-3.0
Open source
Local
Runs on your machine
MCP
Claude Code native
Cross-project
Shared knowledge graph
Pricing

What It Costs

The base Orchestrator and VCT Launcher are free and open source — released, downloadable now. Pro adds RL-scored retrieval reranking at €19/mo. Everything else is coming soon — join the waitlist to be notified.

VibeCoded Orchestrator
Open-source base. Persistent memory and code understanding for Claude Code.
€0
free forever · AGPL-3.0
  • Knowledge graph - always included
  • Code graph - always included
  • 19 agents + 28 skills
  • 20 automation hooks
  • 4 MCP servers
  • VCT Launcher included
Get on GitHub
MultiAgent Orchestrator (MAO)
Multi-agent runtime with hierarchical task planning and a Tauri desktop UI.
Coming Soon
pricing TBD
  • 10 specialist agents on top of Pro
  • Tauri desktop UI
  • Multi-agent maestro runtime
  • Hierarchical task planning
  • All Pro features included
  • No source published (commercial license)
Coming Soon - pricing will be announced at launch
Individual Tools
Standalone desktop tools — Transcrypt, LineArtStudio, ConvertiFacile, SlideMaker. Each app is standalone, runs locally, connects to the orchestrator via the VCT Launcher.
Coming Soon
pricing TBD
  • Transcrypt — local audio transcription
  • LineArtStudio — photo to SVG
  • ConvertiFacile — multi-format converter
  • SlideMaker — PPTX from photos
  • All run locally, no cloud
  • Currently in alpha
Pricing and availability will be finalized at launch
FAQ

Common Questions

The things people ask us most often.

If you got this far, you probably get it.

Join the waitlist. One email at launch. We won't email you weekly.

See what we're building