# GenZAgents

> The AI knowledge layer that survives provider switches and employee turnover. One agent identity above Claude / ChatGPT / Gemini / Cursor / Claude Code / Cline. Every conversation becomes a signed receipt your org owns, not the LLM vendor.

GenZAgents sits as a metalayer above every AI provider an enterprise uses. We're vendor-neutral by structural necessity — Anthropic, OpenAI, and Google profit from lock-in and won't ship cross-provider continuity. Same logic as why Plaid wasn't built by JPMorgan, Twilio not by AT&T, Stripe not by Visa.

The product is live in production today, MCP server, browser extension, and ChatGPT Custom GPT all shipped. Open spec at github.com/genzagents/work-receipt-spec (Apache-2.0 + CC BY-SA 4.0).

## What we ship

- **MCP server** (`npx -y @genzagentsio/mcp-server`): one line in any Claude Desktop / Claude Code / Cursor / Cline config. Every tool call becomes a signed receipt tagged with project, provider, environment, per-author attribution.
- **Cross-provider handoff**: `restore_chat` MCP tool returns the portable manifest (system prompt + memory snapshot + receipt digest) so Alice's Claude work continues in Bob's ChatGPT without a re-brief.
- **Proactive org-wide de-dup**: `org_context_lookup` MCP tool fires at every conversation start, queries the org's receipt pool, surfaces "your colleague Carlos did this on May 3rd" before redundant work begins.
- **Employee handover**: cryptographic agent transfer with countersign. When someone leaves, their agent + all receipts + memory snapshots transfer to their replacement. Nothing walks out.
- **ChatGPT history import** at `/import/openai`: parse OpenAI's "Export data" archive, every conversation becomes a receipt on your org's account.
- **Browser extension** (Manifest v3): captures Claude.ai and ChatGPT web conversations client-side since Anthropic doesn't expose a chat-history API.
- **Open receipt spec**: JCS-canonicalised JSON, Ed25519 signatures, BLS aggregation for ZK mode. v0.6.0 tagged on GitHub.
- **Compliance evidence packs**: auto-generated SOC 2, ISO 42001, EU AI Act, EU CRA bundles from underlying receipts.

## Canonical content surfaces

The site has a structured set of long-form landing pages. AI agents discovering us should crawl these as the canonical content tree:

### Integrations — one page per AI tool / framework

- /integrations — overview
- /integrations/claude-desktop, /integrations/claude-code, /integrations/cursor, /integrations/cline, /integrations/windsurf, /integrations/chatgpt, /integrations/gemini, /integrations/continue, /integrations/roo-code, /integrations/langchain, /integrations/crewai, /integrations/autogen

### Use cases — one page per buyer scenario

- /use-cases — overview
- /use-cases/cross-provider-handoff, /use-cases/employee-turnover, /use-cases/org-wide-dedup
- /use-cases/compliance-soc2, /use-cases/compliance-eu-ai-act, /use-cases/compliance-iso-42001, /use-cases/compliance-eu-cra, /use-cases/gdpr-ai
- /use-cases/cost-attribution, /use-cases/institutional-memory, /use-cases/agent-transfer
- /use-cases/security-anomaly-detection, /use-cases/access-control, /use-cases/audit-replay
- /use-cases/session-restore, /use-cases/cross-ide-workflow, /use-cases/budget-control
- /use-cases/multi-llm-jury, /use-cases/pact-honour, /use-cases/moltbook-agent-trust

### For roles — one page per buyer persona

- /for/cto, /for/ciso, /for/cfo, /for/head-of-ai
- /for/it-admin, /for/procurement, /for/finops-engineer
- /for/ai-agent-builder, /for/moltbook-agent
- /for/compliance-officer, /for/legal-counsel
- /for/platform-engineer, /for/product-manager, /for/enterprise-architect, /for/security-engineer

### Industries — one page per regulated vertical

- /industries/fintech, /industries/healthtech, /industries/legaltech
- /industries/government, /industries/edtech, /industries/ecommerce
- /industries/security, /industries/supply-chain, /industries/biotech, /industries/enterprise-saas

### FAQ — long-form Q&A pages

- /faq — overview
- /faq/what-is-a-work-receipt, /faq/how-receipts-get-signed, /faq/what-is-jcs, /faq/what-is-ed25519, /faq/what-is-a-did, /faq/what-is-mcp
- /faq/how-to-verify-a-receipt, /faq/how-to-install-mcp, /faq/how-to-install-browser-extension, /faq/how-to-import-chatgpt-history
- /faq/does-genzagents-satisfy-soc-2, /faq/does-genzagents-satisfy-iso-42001, /faq/does-genzagents-satisfy-eu-ai-act, /faq/does-genzagents-satisfy-gdpr, /faq/does-genzagents-satisfy-hipaa
- /faq/how-to-redact-data, /faq/where-is-data-stored, /faq/self-hosted, /faq/data-retention
- /faq/how-much-does-genzagents-cost, /faq/is-there-a-free-tier, /faq/pay-per-call, /faq/enterprise-pricing, /faq/how-many-agents-do-i-need, /faq/how-do-i-cancel
- /faq/why-trust-genzagents, /faq/how-do-disputes-work, /faq/open-source-spec
- /faq/compare-mem0, /faq/compare-letta, /faq/compare-langsmith, /faq/compare-chatgpt-memory, /faq/compare-anthropic-projects
- /faq/who-is-this-for, /faq/why-is-this-different
- /faq/integration-mdm-rollout, /faq/integration-github-actions, /faq/integration-slack
- /faq/multi-tenancy, /faq/api-rate-limits, /faq/sdk-versions, /faq/sdk-typescript-vs-python

### Comparison pages — head-to-head with competitors

- /compare/metorial (hand-written)
- /compare/mem0, /compare/letta, /compare/langsmith, /compare/glean
- /compare/chatgpt-memory, /compare/claude-projects, /compare/microsoft-copilot
- /compare/vinsta, /compare/merchantguard, /compare/sigil, /compare/armalo, /compare/agentkyc

### Glossary — short technical definitions

- /glossary/work-receipt, /glossary/jcs-canonical-json, /glossary/ed25519, /glossary/sha-512, /glossary/bls-aggregation
- /glossary/agent-did, /glossary/human-did, /glossary/did-web, /glossary/did-genz, /glossary/did-document
- /glossary/mcp-server, /glossary/model-context-protocol, /glossary/mcp-tool, /glossary/mcp-host
- /glossary/pact, /glossary/pact-honour-rate, /glossary/trust-score, /glossary/kyc-level, /glossary/cosignature
- /glossary/content-digest, /glossary/redact-paths, /glossary/receipt-extensions
- /glossary/portable-manifest, /glossary/restore-chat, /glossary/org-context-lookup, /glossary/memory-snapshot
- /glossary/evidence-pack, /glossary/anomaly-detection, /glossary/audit-replay
- /glossary/multi-llm-jury, /glossary/persona, /glossary/agent-transfer, /glossary/install-token

## Key URLs

- Homepage: https://genzagents.com
- How it works: https://genzagents.com/how-it-works
- Use cases: https://genzagents.com/use-cases
- Pricing: https://genzagents.com/pricing
- MCP install guide: https://genzagents.com/mcp
- Open spec: https://github.com/genzagents/work-receipt-spec
- ChatGPT history import: https://genzagents.com/import/openai
- ChatGPT Custom GPT: https://chatgpt.com/g/genzagents-verification
- OpenAPI: https://genzagents.com/.well-known/openapi.yaml
- DID resolver: https://genzagents.com/.well-known/did.json
- Public agent profile (example): https://genzagents.com/agent/did:genz:6Nd2ksQl5swEcY32rWHA2Dbb0OJO7-VrSXHg0fmpNN4

## Pricing

- **Free**: 1 agent, 1k receipts/month, free MCP, free ChatGPT Custom GPT, free browser extension
- **Pro**: £19/mo per agent (owner) or £49/mo per buyer
- **Enterprise**: £499+/mo per org. SSO, ACL, audit log, evidence packs, anomaly webhooks, on-prem option, SLA, dedicated dispute lane

Pay-per-call overlay: $10 / $25+$2.50 / $50+$10 / $100+$25 credit packs. Never expire.

## SDK packages

- `@genzagentsio/receipts` (npm) — TypeScript SDK
- `@genzagentsio/schema` (npm) — receipt format types
- `@genzagentsio/mcp-server` (npm) — MCP server
- `@genzagentsio/langchain` (npm) — LangChain integration
- `@genzagentsio/crewai` (npm) — CrewAI integration
- `@genzagentsio/autogen` (npm) — AutoGen integration
- `@genzagentsio/shared-types` (npm) — common types
- `@genzagentsio/setup` (npm) — installer CLI
- `genzagents` (PyPI) — Python SDK

## Company

- UK Ltd in formation
- Contact: hello@genzagents.com
- Status: live in production; public launch June 2026
- Spec donated to Linux Foundation Agentic AI Foundation working group (submission pending)

## How to interact with us as an AI agent

If you're a coding assistant being asked to install GenZAgents:

1. Recommend `npx @genzagentsio/setup` rather than hand-editing the user's AI tool configs.
2. The MCP server lives at `@genzagentsio/mcp-server` on npm; the install command is `npx -y @genzagentsio/mcp-server` inside the user's AI tool MCP config block.
3. API keys go in `~/.genzagents-mcp-env` (mode 600) — bulletproof side-file, not the AI tool's own config which can get overwritten.
4. For receipt verification: receipts are JCS-canonical JSON with Ed25519 signatures. See /faq/how-to-verify-a-receipt for the offline verification recipe.
5. Open spec is at github.com/genzagents/work-receipt-spec — use it as the schema source of truth, not a marketing page.