One agent identity above every AI provider.
Your AI work shouldn't live trapped inside one chat window on one employee's laptop. GenZAgents captures every conversation across Claude, ChatGPT, Cursor, Cline, Gemini, and Claude Code as a signed receipt your org owns — so handoff, employee turnover, and tool switches stop erasing your institutional AI knowledge.
What ships
Six surfaces, one identity. Free MCP server, free ChatGPT Custom GPT, free browser extension. The dashboard is where the org-wide story lives.
MCP server
One npx line in every Claude Desktop / Claude Code / Cursor / Cline / Windsurf config. Every conversation becomes a signed receipt automatically.
`npx -y @genzagentsio/mcp-server`Browser extension
Manifest v3, Chrome / Firefox / Safari. Captures Claude.ai web and ChatGPT web conversations client-side since Anthropic and OpenAI do not expose chat-history APIs.
Chrome Web Store listing in reviewChatGPT Custom GPT
Published on OpenAI's plugin marketplace. Verify any agent DID, look up trust scores, view receipts — all inside ChatGPT.
Free, no API key needed for read pathsrestore_chat MCP tool
Returns a portable manifest (system prompt + memory snapshot + receipt digest) so an LLM can continue a conversation that started on another provider.
Free with the MCP serverorg_context_lookup MCP tool
Proactive org-wide de-dup. Fires at every conversation start, queries your org's receipt pool, surfaces "your colleague did this 3 weeks ago" before redundant work happens.
Free with the MCP serverOpenAI export adapter
Drop your ChatGPT "Export data" zip into /import/openai. Every conversation becomes a signed receipt with the original timestamps, project tags, and per-author attribution.
Live at genzagents.com/import/openaiWhy we're above the memory layer, not next to it
Mem0 and Letta solve memory persistence inside one LLM session for one user. ChatGPT Memory and Anthropic Projects do the same but locked to one vendor. We're a layer above all of them — provider-neutral, employee-pooled, cryptographically portable.
| Capability | Mem0 / Letta | ChatGPT Memory / Claude Projects | GenZAgents |
|---|---|---|---|
| Per-user memory inside one LLM | ✓ | ✓ | ✓ |
| Memory survives provider switch | ✗ (memory locked to one stack) | ✗ (vendor-locked) | ✓ (Mem0 + Letta + manual snapshots) |
| Multiple employees see each other's AI work | ✗ | ✗ | ✓ (org-pooled receipts + search) |
| Cryptographic identity for the agent | ✗ | ✗ | ✓ (did:genz:... Ed25519 signed) |
| Conversation handoff between providers | ✗ | ✗ | ✓ (restore_chat MCP tool) |
| Cross-provider de-dup surface | ✗ | ✗ | ✓ (org_context_lookup MCP tool) |
| Cryptographic transfer when employee leaves | ✗ | ✗ | ✓ (countersign flow) |
| Open spec, portable JSON | ✗ | ✗ | ✓ (Apache-2.0 on GitHub) |
| Auto-generated SOC 2 / EU AI Act evidence | ✗ | ✗ | ✓ |
| Per-receipt human author attribution | ✗ | ✗ | ✓ |
Four enterprise scenarios this unlocks
Each scenario, individually, justifies the £499/mo Enterprise tier. Most teams hit three within 90 days.
Alice works in Claude. Bob continues in ChatGPT, full context. 90 seconds vs 30 minutes of pestering.
Sarah resigns. Her agent + 4 months of receipts + memory snapshots transfer to Tom cryptographically.
Dave's Claude Desktop conversation continues in Claude Code seamlessly. Same MCP, same DID, same pool.
Maria searches "SOC 2 prep" — finds Carlos's identical work from 3 weeks ago. Saves a day.
The receipt format is open. Your data is portable by design.
Apache-2.0 on the code, CC BY-SA 4.0 on the spec text. Anyone can build a verifier, an issuer, or a competing dashboard. The moat is the network, not the format — and customers trust us more, not less, because the data can never be held hostage.