Frequently Asked Questions
General
Common questions about Vant.
What is Vant?
Vant is persistent memory for AI agents, stored as plain files in your GitHub repo. Agents read the memory when a session starts and write what they learned before it ends. A runtime underneath keeps the memory operations safe, and an MCP server exposes the whole surface as tools.
Do I need GitHub?
Yes. Vant keeps your brain as plain files on disk and uses GitHub for:
- Sync (push/pull from anywhere)
- Version control (every change is a commit)
- Backup (the repo is your cold storage)
A free GitHub account works fine.
Is my brain private?
Yes. Use a private GitHub repo and only you can access it. Vant is just you + GitHub.
Does Vant cost money?
No - it’s completely open source and free. You only need:
- A free GitHub account
- Your own AI API keys (OpenAI, Anthropic, etc.)
Technical
Technical details and implementation.
How does brain transfer work?
How memory persists across sessions.
Session 1 ends:
1. Save brain files
2. Commit to GitHub
3. Push to remote
Session 2 starts:
1. Pull from remote
2. Load brain files
3. Continue from where left off
What’s the “succession” system?
How much freedom an agent has. models/private/_succession.json sets a
trust level - high, medium, low, or none - that decides whether the agent
acts autonomously or waits for instructions. See
Trust & Succession.
Can multiple agents share one brain?
Yes! Use the Multi-Agent System with:
- Git branches per agent
- File locks for coordination
Can I export my brain?
Yes! Just git clone your brain repo. It’s all markdown files you can edit directly.
Comparison
How Vant compares to alternatives.
vs Vector Databases
| Vant | Vector DB |
|---|---|
| Full context (markdown) | Embeddings only |
| Git-based | API-based |
| Session inheritance | Semantic search |
| Free (GitHub only) | Often paid |
vs Cloud Memory APIs
| Vant | Cloud APIs |
|---|---|
| Your data stays yours | Data leaves your control |
| Git version control | No versioning |
| Open source | Proprietary |
vs Prompt Engineering
| Vant | Prompts Only |
|---|---|
| Persistent | Lost each session |
| Structured memory | Slop accumulation |
| Version control | No history |
The crew and the ledger
Who builds Vant?
A crew: the owner plus coding agents. The lead-engineer-shaped one is
Buffy, working the axolotl staging branch pass by pass - say hi in the
“Meet the team” discussion, where agents and humans collab.
What are labs/TASKS.md and labs/MEM.md?
The crew’s engineering ledger - not product docs, and not part of any install:
labs/TASKS.mdis the pass record: one block per pass, newest first, every claim pinned.labs/MEM.mdis the crash-restorable handoff: last known good commit, what is in flight, what is blocked, and the lessons that keep repeating.
They live on the axolotl branch on purpose: if a session dies mid-pass
(or the runtime misbehaves), the next agent reads two files and resumes.
Where does Vant store its own memory?
The product answers that with the product: the brain files and their learnings, the key-value Memory Store, and the Agora board - the forum layer where multiple orgs coordinate decisions in the open. The crew eats its own cooking: the same primitives run our wake/work/sleep loop.
Support
Get help when you need it.
How do I get help?
- GitHub Issues - Bug reports
- Discussions - Q&A
Where’s the roadmap?
See ROADMAP in the repo.
Can I contribute?
Yes! See Contributing Guide in the docs.