Frequently Asked Questions

vem is a project memory system for AI agents and developers. It keeps a local workspace cache in .vem/ and syncs canonical memory — tasks, decisions, and snapshots — to the cloud. Agents access memory via MCP tools; developers use the CLI. Every snapshot is verified against Git metadata so context is always auditable and trustworthy.

Yes! vem exposes a standard API that can be easily integrated with LangChain, Vercel AI SDK, and other popular frameworks.

Absolutely. All data is encrypted at rest and in transit. We use Row Level Security to ensure strict isolation between organizations. Snapshot payloads are also scanned for secrets before they are stored.

AI credits power intelligent features in vem — semantic search, memory analysis, and implementation planning. Each plan includes a monthly credit allowance: Free gets 500, Pro gets 5,000, and Ultra gets 25,000. Credits are consumed at roughly 1 credit per 1,000 tokens processed. If you need more, you can top up with one-time credit packs that never expire.

AI-powered features will return a clear error until your credits reset at the start of your next billing cycle. You can purchase a credit pack any time to top up immediately — credits stack with your monthly allowance.

Git provides a built-in audit trail. vem keeps .vem/ gitignored — it never pollutes your repo — but uses Git metadata (commit hashes) to verify cloud snapshots. When you push a snapshot and then push a commit, vem matches the git_hash to finalize the snapshot as verified. This gives you cryptographic provenance without storing anything in your repo.

Truth Drift occurs when an agent or developer marks a task as complete without providing verifiable evidence (like a commit hash or test output). vem detects this "drift" and flags the memory as suspicious to maintain high data integrity.

We offer on-premise deployments for Enterprise customers. Contact sales for more information.