Integrations › Claude Code
Trusted State for Claude Code: richest conversation capture + handoff to Cursor.
Get API Key →Claude Code is powerful but loses continuity between sessions. It re-reads your codebase every time. It doesn't remember your preferences, your architecture decisions, or the bugs it already fixed. CLAUDE.md helps, but it's manual, unversioned, and doesn't hand off to your other tools.
Context survives session resets.
Same project. Different day. Memory intact.
One command — recommended:
npx @verifiedstate/sync init
# Anonymous key + MCP + daemon capture + hot-context.
# Claim a durable account later: vsync claim
# Capture: Claude JSONL (full conversation when available)
# Retrieval: hot-context auto-refresh + MCP session_load (agent protocol)
# Advanced MCP-only:
claude mcp add verifiedstate --transport http \
https://mcp.verifiedstate.ai/mcp{
"receipt_id": "rct_7f3a...",
"assertion": "User prefers dark mode",
"status": "verified",
"confidence": 0.94,
"signature": "ed25519:a8c3f1...",
"verified_at": "2026-04-06T14:32:01Z",
"chain_position": 42
}| Feature | Without | With VerifiedState |
|---|---|---|
| Between sessions | Re-reads entire codebase | Loads verified context instantly |
| Architecture decisions | Lost when session ends | Persisted and available to connected agents |
| Contradictions | Agent contradicts yesterday's work | Conflicts detected automatically |
| Capture quality | Nothing persisted | Full conversation when Claude capture is rich |
It complements it. vsync manages a VerifiedState section in CLAUDE.md and syncs hot-context. Your own CLAUDE.md content is preserved.
No. Install once, work normally. Open VerifiedState when you need History, Search, Now, or Review.
Any version that supports MCP servers.
Usually no. Run init in the project, then keep working. Verify with vsync status.
Add VerifiedState once. Keep project state available across your AI agents.