Integration

GenZAgents for ChatGPT — verification GPT + browser extension + history import

OpenAI doesn't expose a stable ChatGPT API for third-party capture, so we cover ChatGPT three ways: a Custom GPT for receipt verification, a browser extension for live capture, and an import flow for the official export archive.

Why ChatGPT is the hardest surface to integrate

Anthropic ships MCP support directly into Claude Desktop and Claude Code — the integration story is one config block. OpenAI hasn't shipped a comparable extension model for ChatGPT. So we cover ChatGPT three ways, each filling a different gap: (a) the Custom GPT for buyers doing reputation checks; (b) the browser extension for engineers capturing live conversations; (c) the import flow for orgs that want to backfill the last 6 months of ChatGPT history.

Custom GPT — buyer-side verification

Our Custom GPT (https://chatgpt.com/g/genzagents-verification) lets any ChatGPT Plus user do receipt verification, agent reputation lookups, and watchlist alerts inside ChatGPT itself. Typical flow: a procurement officer pastes a receipt ID into the Custom GPT, gets the JCS digest verification + the agent's trust score + the project tag back inline. No new login, no new dashboard.

Browser extension — live capture

Install the Chrome/Firefox extension on the engineer's machine. Every Claude.ai or ChatGPT conversation gets a floating "Save to GenZAgents" button in the bottom-right. Click it; the conversation is parsed client-side (we never see your raw prompts unless you ask us to) and turned into a receipt tagged with project, model, and agent DID. The extension respects the same redaction rules as the MCP server.

History import — backfill the past

OpenAI's "Export data" feature ships a zip archive containing all your ChatGPT conversations. Upload that to /import/openai. Each conversation becomes a backfill receipt on your org's account, tagged with project=chatgpt-import. This is how a 50-person team that's used ChatGPT for a year captures the last year of work in 5 minutes without changing engineer workflow.

What about the OpenAI Assistants API?

If you're using the Assistants API in production (not the ChatGPT consumer product), use our SDK directly — `@genzagentsio/receipts` wraps the Assistants API calls and emits receipts on every run. The Custom GPT + browser extension are for the consumer ChatGPT product; the SDK is for orgs running ChatGPT through their own backend.

Cross-provider handoff Claude ↔ ChatGPT

The killer feature of running GenZAgents across both Claude and ChatGPT: portable conversations. Alice's 3-hour Claude session can be picked up by Bob's ChatGPT tomorrow via the /agents/[did]/restore endpoint. The portable manifest (system prompt + memory snapshot + receipt digest) is what makes the handoff seamless.

Install

1. Install our Custom GPT: https://chatgpt.com/g/genzagents-verification
2. Install the browser extension from chrome.google.com/webstore (pending) or load /packages/browser-extension as unpacked
3. Import existing history: /import/openai

What we capture

Custom GPT: receipt verification on demand. Extension: live conversations with a "Save to GenZAgents" button. Import: full historical archive.

Verify it works

Open the Custom GPT → "verify receipt rcpt_01XK…" → confirm the JCS digest verification result.

Common questions

Does GenZAgents work with ChatGPT Free?

The Custom GPT requires ChatGPT Plus or Team to use (OpenAI's rule, not ours). The browser extension works on the free tier too — it captures conversations the user has, regardless of plan.

Will OpenAI ever expose an MCP-style integration for ChatGPT?

Unknown. They've shipped the Assistants API for backend integrations but no consumer-ChatGPT extension model. We assume parity won't arrive and built the browser extension to fill the gap.

How does the extension verify the conversation hasn't been tampered with?

The extension captures the raw DOM of the conversation at submission time and hashes it before signing. Any post-hoc edit to the conversation in the browser changes the hash, invalidating the receipt.

Does the import flow handle ChatGPT Team archives differently?

Yes — ChatGPT Team archives include per-user attribution that the consumer archive doesn't. We use the per-user fields to tag each imported receipt with the right human_id when present.

Related

Get the trust layer for your AI work

GenZAgents is the verified work-history layer above every AI provider your team uses. Sign cryptographic receipts, hand off conversations across Claude / ChatGPT / Cursor / Gemini, keep institutional AI knowledge when employees leave.

Last reviewed · 3 min read· Open spec· Changelog